WO1997036376A1 - Table-based compression with embedded coding - Google Patents
Table-based compression with embedded coding Download PDFInfo
- Publication number
- WO1997036376A1 WO1997036376A1 PCT/US1997/004879 US9704879W WO9736376A1 WO 1997036376 A1 WO1997036376 A1 WO 1997036376A1 US 9704879 W US9704879 W US 9704879W WO 9736376 A1 WO9736376 A1 WO 9736376A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- vectors
- codebook
- image
- stage
- vector
- Prior art date
Links
Classifications
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
- H03M7/30—Compression; Expansion; Suppression of unnecessary data, e.g. redundancy reduction
- H03M7/3082—Vector coding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
- H04N19/146—Data rate or code amount at the encoder output
- H04N19/147—Data rate or code amount at the encoder output according to rate distortion criteria
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/189—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the adaptation method, adaptation tool or adaptation type used for the adaptive coding
- H04N19/19—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the adaptation method, adaptation tool or adaptation type used for the adaptive coding using optimisation based on Lagrange multipliers
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/48—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using compressed domain processing techniques other than decoding, e.g. modification of transform coefficients, variable length coding [VLC] data or run-length data
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/593—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial prediction techniques
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
- H04N19/61—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding in combination with predictive coding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
- H04N19/63—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding using sub-band based transform, e.g. wavelets
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/90—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using coding techniques not provided for in groups H04N19/10-H04N19/85, e.g. fractals
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/90—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using coding techniques not provided for in groups H04N19/10-H04N19/85, e.g. fractals
- H04N19/94—Vector quantisation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/90—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using coding techniques not provided for in groups H04N19/10-H04N19/85, e.g. fractals
- H04N19/96—Tree coding, e.g. quad-tree coding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/115—Selection of the code volume for a coding unit prior to coding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
- H04N19/146—Data rate or code amount at the encoder output
Definitions
- the present invention relates to data processing and, more particularly, to data compression, for example as applied to still and video images, speech and music.
- a major objective of the present invention is to enhance collaborative video applications over heterogeneous networks of inexpensive general purpose computers.
- Effectiveness can be measured in terms of the amount of distortion resulting for a given degree of compression.
- the distortion can be expressed in terms of the square of the difference between corresponding pixels averaged over the image, i.e., mean square error (less is better).
- the mean square error can be: 1) weighted, for example, to take variations in perceptual sensitivity into account; or 2) unweighted.
- the extent of compression can be measured either as a compression ratio or a bit rate.
- the compression ratio (more is better) is the number of bits of an input value divided by the number of bits in the expression of that value in the compressed code (averaged over a large number of input values if the code is variable length).
- the bit rate is the number of bits of compressed code required to represent an input value. Compression effectiveness can be characterized by a plot of distortion as a function of bit rate.
- lossless compression techniques tend to be limited to compression ratios of about 2, whereas compression ratios of 20 to 500 are desired for collaborative video applications. Lossy compression techniques always result in some distortion. However, the distortion can be acceptable, even imperceptible, while much greater compression is achieved.
- Collaborative video is desired for communication between general purpose computers over heterogeneous networks, including analog phone lines, digital phone lines, and local-area networks. Encoding and decoding are often computationally intensive and thus can introduce latencies or bottlenecks in the data stream. Often dedicated hardware is required to accelerate encoding and decoding. However, requiring dedicated hardware greatly reduces the market for collaborative video applications. For collaborative video, fast, software-based compression would be highly desirable.
- Heterogeneous networks of general purpose computers present a wide range of channel capacities and decoding capabilities.
- One approach would be to compress image data more than once and to different degrees for the different channels and computers.
- this is burdensome on the encoding end and provides no flexibility for different computing power on the receiving end.
- a better solution is to compress image data into a low-compression/low distortion code that is readily scalable to greater compression at the expense of greater distortion.
- the lossy compression techniques practically required for video compression generally involve quantization applied to monochrome (gray-scale or color component) images.
- quantization a high-precision image description is converted to a low-precision image description, typically through a many -to-one mapping.
- Quantization techniques can be divided into scalar quantization (SQ) techniques and vector quantization (VQ) techniques. While scalars can be considered one-dimensional vectors, there are important qualitative distinctions between the two quantization techniques.
- Vector quantization can be used to process an image in blocks, which are represented as vectors in an n-dimensional space. In most monochrome photographic images, adjacent pixels are likely to be close in intensity. Vector quantization can take advantage of this fact by assigning more representative vectors to regions of the n-dimensional space in which adjacent pixels are close in intensity than to regions of the n-dimensional space in which adjacent pixels are very different in intensity. In a comparable scalar quantization scheme, each pixel would be compressed independently; no advantage is taken of the correlations between adjacent pixels. While, scalar quantization techniques can be modified at the expense of additional computations to take advantage of correlations, comparable modifications can be applied to vector quantization. Overall, vector quantization provides for more effective compression than does scalar quantization.
- the compressed data can include reduced precision expressions of the representative values. Such a representation can be readily scaled simply by removing one or more least-significant bits from the representative value.
- the representative values are represented by indices; however, scaling can still take advantage of the fact that the representative values have a given order in a metric dimension.
- representative vectors are distributed in an n-dimensional space. Where n>l, there is no natural order to the representative vectors. Accordingly, they are assigned effectively arbitrary indices. There is no simple and effective way to manipulate these indices to make the compression scalable.
- vector quantization The final distinction between vector and scalar quantization is more quantitative than qualitative.
- the computations required for quantization scale dramatically (more than linearly) with the number of pixels involved in a computation.
- scalar quantization one pixel is processed at a time.
- vector quantization plural pixels are processed at once.
- the number of pixels processed at once becomes 16 and 64, respectively.
- full-search vector quantization computes the distances in an n-dimensional space of an image vector from each representative vector Accordingly, vector quantization tends to be much slower than scalar quantization and, therefore, limited to off-line compression applications.
- Tree-structured VQ Comparisons are performed in pairs. For example, the first two measurements can involve codebook points in symmetrical positions in the upper and the Iower halves of a vector space. If an image input vector is closer to the upper codebook point, no further comparisons with codebook points in the lower half of the space are performed. Tree-structured VQ works best when the codebook has certain symmetries. However, requiring these symmetries reduces the flexibility of codebook design so that the resulting codebook is not optimal for minimizing distortion. Furthermore, while reduced, the computations required by tree-structured VQ can be excessive for collaborative video applications.
- TBVQ table-based vector quantization
- vector quantization is typically preceded by a transform such as a wavelet transform.
- the present invention provides for data compression using a hierarchical table implementing a block transform and outputting a variable-rate, embedded code.
- a hierarchical table implementing a block transform and outputting a variable-rate, embedded code.
- a counterintuitive aspect of the present invention is the inco ⁇ oration of a codebook of a type used for structured vector quantization in a compression table.
- Structured vector quantization is designed to reduce the computations required for compression while accepting a small increase in distortion relative to full-search vector quantization.
- this tradeoff is a poor one in the context of tables, since all the computations are pre-computed.
- a codebook design procedure used for tree-structured vector quantization is used, not to reduce computations, but to provide a codebook that can be mapped readily to an embedded code.
- bits are arranged in order of significance. When the least significant bit of a multi-bit index to a first codebook vector is dropped, the result is an index of a codebook vector near the first codebook vector.
- an embedded code is readily scaled to provide a variable-rate system.
- An embedded code can readily be made variable length to minimize entropy and reduce the bit rate for a net gain in compression effectiveness.
- any loss of effectiveness resulting from the use of a structured vector quantization codebook is at least partially offset by the gain in compression effectiveness resulting from the use of a variable-length code.
- Block transforms can express data so that information can be separated by significance. This makes it feasible to apply more compression to less significant data for a net gain in the apparent effectiveness of the compression.
- a perceptually weighted proximity measure can be used during codebook design.
- an unweighted or less perceptually weighted proximity measure should be used during a table fill-in procedure to minimize distortion.
- a further aspect of the invention is the inco ⁇ oration of considerations other than perceptually weighted or unweighted proximity measures in codebook design.
- entropy constraints can be imposed on codebook design to enhance bit rate.
- a joint entropy and distortion measure can be used to select nodes to be grown or pruned. If the joint measure is applied on a node-by-node basis, virtually continuous scalability can be provided while maintaining high compression effectiveness at each available bit rate.
- a final aspect of the invention takes advantage of the lower memory requirements afforded by hierarchical tables. Hierarchical tables raise the issue of how to inco ⁇ orate structures, constraints, and transforms in a table. In the case of the block transforms, the transforms are used in codebook design at every stage of the table. However, in the case of structures and constraints used to provide variable-length codes, these are best restricted to design of the last-stage table only.
- FIGURE 1 is a schematic illustration of an image compression system in accordance with the invention.
- FIGURE 2 is a flow chart for designing the compression system of FIG. 1 in accordance with the present invention.
- FIGURE 3 is a schematic illustration of a decision tree for designing an embedded code for the system of FIG. 1.
- FIGURE 4 is a graph indicating the performance of the system of FIG. 1.
- FIGURES 5-8 are graphs indicating the performance of other embodiments of the present invention.
- an image compression system Al comp ⁇ ses an encoder ENC, communications lines LAN, POTS, and IDSN, and a decoder DEC, as shown m FIG. 1.
- Encoder ENC is designed to compress an original image for distribution over the communications lines.
- Communications lines POTS, IDSN, and LAN differ widely in bandwidth.
- "Plain Old Telephone Service” line POTS which includes an associated modem, conveys data at a nominal rate of 28.8 kilobaud (symbols per second).
- "Integrated Data Services Network” line IDSN conveys data an order of magnitude faster.
- "Local Area Network” line LAN conveys data at about 10 megabits per second.
- Many receiving and decoding computers are connected to each line, but only one computer is represented in FIG. 1 by decoder DEC. These computers decompress the transmission from encoder ENC and generate a reconstructed image that is faithful to the original image.
- Encoder ENC comprises a vectorizer VEC and a hierarchical lookup table HLT, as shown in FIG. 1.
- Vectorizer VEC converts a digital image into a series of image vectors Ii.
- Hierarchical lookup table HLT converts the series of vectors Ii into three series of indices ZAi, ZBi, and ZCi.
- Index ZAi is a high-average-precision variable-length embedded code for transmission along line LAN
- index ZBi is a moderate-average-precision variable-length embedded code for transmission along line IDSN
- index ZCi is a low-average-precision variable-length embedded code for transmission along line POTS.
- the varying precision accommodates the varying bandwidths of the lines.
- Vectorizer VEC effectively divides an image into blocks Bi of 4x4 pixels, where i is a block index varying from 1 to the total number of blocks in the image. If the original image is not evenly divisible by the chosen block size, additional pixels can be added to sides of the image to make the division even in a manner known in the art of image analysis.
- Each vector element Vj is expressed in a suitable precision, e.g., eight bits, representing a monochromatic (color or gray scale) intensity associated with the respective pixel.
- Vectorizer VEC presents vector elements Vj to hierarchical lookup table HLT in adjacently numbered odd- even pairs (e.g., VI, V2) as shown in FIG. 1.
- Hierarchical lookup table HLT includes four stages S I , S2, S3, and S4. Stages S I , S2, and S3 collectively constitute a preliminary section PRE of hierarchical lookup table HLT, while fourth stage S4 constitutes a final section.
- the tables of the preliminary section stages SI, S2, and S3 are shown multiple times to represent the number of times they are used per image vector. For example, table Tl receives eight pairs of image vector elements Vj and outputs eight respective first-stage indices Wj. If the processing power is affordable, a stage can include several tables of the same design so that the pairs of input values can be processed in parallel.
- the pu ⁇ ose of preliminary section PRE is to reduce the number of possible vectors that must be compressed with minimal loss of perceptually relevant information.
- the pu ⁇ ose of final- stage table T4 is to map the reduced number of vectors many-to-one to each set of embedded indices.
- Table T4 has 2 20 entries corresponding to the concatenation of two ten-bit inputs.
- Tables T2, and T3 are the same size as table T4, while table Tl is smaller with 2 16 entries.
- the total number of addresses for all stages of hierarchical vector table HLT is less than four million, which is a practical number of table entries. For computers where that is excessive, all tables can be limited to 2 16 entries, so that the total number of table entries is about one million.
- Each preliminary stage table Tl, T2, T3, has two inputs and one output, while final stage
- T4 has two inputs and three outputs. Pairs of image vector elements Vj serve as inputs to first stage table T 1.
- the vector elements can represent values associated with respective pixels of an image block. However, the invention applies as well if the vector elements Vj represent an array of values obtained after a transformation on an image block.
- the vector elements can be coefficients of a discrete cosine transform applied to an image block.
- each input vector is in the pixel domain and hierarchical table HLT implements a discrete cosine transform.
- each vector value Vj is treated as representing a monochrome intensity value for a respective pixel of the associated image block, while indices Wj, Xj, Yj, ZA, ZB, and ZC, represent vectors in the spatial frequency domain.
- Each pair of vector values (Vj, V(j+1)) represents with a total of sixteen bits a 2x1 (column x row) block of pixels.
- (VI, V2) represents the 2x1 block highlighted in the leftmost replica of table Tl in FIG. 1.
- Table Tl maps pairs of vector element values many -to- one to eight-bit first-stage indices Wj; in this case, j ranges from 1 to 8.
- Each eight-bit Wj also represents a 2x1 -pixel block. However, the precision is reduced from sixteen bits to eight bits.
- the eight first-stage indices Wj are combined into four adjacent odd-even second-stage input pairs; each pair (Wj, W(j+1)) represents in sixteen-bit precision the 2x2 block constituted by the two 2x1 blocks represented by the individual first-stage indices Wj.
- (W1 ,W2) represents the 2x2 block highlighted in the leftmost replica of table T2 in FIG. 1.
- Second stage table T2 maps each second-stage input pair of first-stage indices many-to-one to a second stage index Xj.
- the eight first-stage indices yield four second-stage indices XI , X2, X3, and X4.
- Each of the second stage indices Xj represents a 2x2 image block with eight-bit precision.
- the four second-stage indices Xj are combined into two third-stage input pairs (XI , X2) and (X3,X4), each representing a 4x2 image block with sixteen-bit precision.
- (XI , X2) presents the upper half block highlighted in the left replica of table T3, while (X3,X4) represents the lower half block highlighted in the right replica of table T3 in FIG. 1.
- Third stage table T3 maps each third-stage input pair many-to-one to eight-bit third-stage indices Yl and Y2. These two indices Yl and Y2 are the output of preliminary section PRE in response to a single image vector.
- fourth-stage table T4 maps fourth- stage input pairs many-to-one to each of the embedded indices ZA, ZB, and ZC. For an entire image, there are many image vectors Ii, each yielding three respective output indices ZAi, ZBi, and ZCi. The specific relationship between inputs and outputs is shown in Table I below as well as in FIG. 1.
- Decoder DEC is designed for decompressing an image received from encoder ENC over a LAN line.
- Decoder DEC includes a code pruner 51, a decode table 52, and an image assembler 53.
- Code pruner 51 performs on the receiving end the function that the multiple outputs from stage S4 perform on the transmitting end: allowing a tradeoff between fidelity and bit rate.
- Code pruner 51 embodies the criteria for pruning index ZA to obtain indices ZB and ZC; alternatively, code pruner 51 can pass index ZA unpruned.
- the code pruning effectively reverts to an earlier version of the greedily grown tree.
- the pruned codes generated by a code pruner need not match those generated by the encoder.
- the code pruner could provide a larger set of alternatives.
- the pruning function can merely involve dropping a fixed number of least-significant bits from the code. This truncation can take place at the encoder at the hierarchical table output and/or at the decoder.
- a more sophisticated approach is to prune selectively based on an entropy constraint.
- Decode table 52 is a lookup table that converts codes to reconstruction vectors. Since the code indices represent codebook vectors in a spatial frequency domain, decode table 52 implements a pre-computed inverse discrete cosine transform so that the reconstruction vectors are in a pixel domain. Image assembler 53 converts the reconstruction vectors into blocks and assembles the reconstructed image from the blocks.
- decoder DEC is implemented in software on a receiving computer.
- the software allows the fidelity versus bit rate tradeoff to be selected.
- the software then sets code pruner 51 according to the selected code precision.
- the software includes separate tables for each setting of code pruner 51. On the table corresponding to the current setting of code pruner 51 is loaded into fast memory (RAM).
- fast memory RAM
- lookup table 52 is smaller when pruning is activated.
- the pruning function allows fast memory to be conserved to match: 1) the capacity of the receiving computer; or 2) the allotment of local memory to the decoding function.
- a table design method Ml is executed for each stage of hierarchical lookup table HLT, with some variations depending on whether the stage is the first stage SI, an intermediate stage S2, S3, or the final stage S4.
- method Ml includes a codebook design procedure 10 and a table fill-in procedure 20.
- fill-in procedure 20 must be preceded by the respective codebook design procedure 10.
- table T3 can be filled in before the codebook for table T2 is designed.
- codebook design procedure 10 begins with the selection of training images at step 1 1.
- the training images are selected to be representative of the type or types of images to be compressed by system Al . If system Al is used for general pu ⁇ ose image compression, the selection of training images can be quite diverse. If system Al is used for a specific type of image, e.g., line drawings or photos, then the training images can be a selection of images of that type. A less diverse set of training images allows more faithful image reproduction for images that are well matched to the training set, but less faithful image reproduction for images that are not well matched to the training set.
- the training images are divided into 2x1 blocks, which are represented by two- dimensional vectors (Vj,V(J+l)) in a spatial pixel domain at step 12. For each of these vectors Vj characterizes the intensity of the left pixel of the 2x1 block and V(J+1) characterizes the intensity of the right pixel of the 2x1 block.
- codebook design and table fill in are conducted in the spatial pixel domain.
- steps 13, 23, 25 are not executed for any of the stages.
- a problem with the pixel domain is that the terms of the vector are of equal importance: there is no reason to favor the intensity of the left pixel over the intensity of the right pixel, and vice versa.
- table Tl to reduce data while preserving as much information relevant to classification as possible, it is important to express the information so that more important information is expressed independently of less important information.
- a discrete cosine transform is applied at step 13 to convert the two-dimensional vectors in the pixel domain into two-dimensional vectors in a spatial frequency domain. The first value of this vector corresponds to the average intensities of the left and the right pixels, while the second value of the vector corresponds to the difference in intensities between the left and the right pixels.
- the codebook is designed at step 14.
- step 14 would determine the set of 1024 vectors that would yield the minimum distortion for images having the expected probability distribution of 2x1 input vectors. While the problem of finding the ideal codebook vectors can be formulated, it cannot be solved generally by numerical methods. However, there is an iterative procedure that converges from an essentially arbitrary set of "seed" vectors toward a "good” set of codebook vectors. This procedure is known alternatively as the “cluster compression algorithm", the "Linde-Buzo-Gray” algorithm, and the “generalized Lloyd algorithm” (GLA).
- cluster compression algorithm the "Linde-Buzo-Gray” algorithm
- GLA Generalized Lloyd algorithm
- the procedure begins with a set of seed vectors.
- the training set of 2x1 spatial frequency vectors generated from the training images are assigned to the seed vectors on a proximity basis. This assignment defines clusters of training vectors around each of the seed vectors.
- the weighted mean vector for each cluster replaces the respective seed vector.
- the mean vectors provide better distortion performance than the seed vectors; a first distortion value is determined for these first mean vectors.
- Further improvement is achieved by re-clustering the training vectors around the previously determined mean vectors on a proximity basis, and then finding new mean vectors for the clusters.
- This process yields a second distortion value less than the first distortion value.
- the difference between the first and second distortion values is the first distortion reduction value.
- the process can be iterated to achieve successive distortion values and distortion reduction values.
- the distortion values and the distortion reduction values progressively diminish. In generally, the distortion reduction value does not reach zero. Instead, the iterations can be stopped with the distortion reduction values fall below a predetermined threshold--/, e., when further improvements in distortion are not worth the computational effort.
- This splitting technique begins by determining a mean for the set of training vectors. This can be considered the result of applying a single GLA iteration to a single arbitrary seed vector as though the codebook of interest were to have one vector.
- the mean vector is perturbed to yield a second "perturbed" vector.
- the mean and perturbed vectors serve as the two seed vectors for the next iteration of the splitting technique.
- the perturbation is selected to guarantee that some training vectors will be assigned to each of the two seed vectors.
- the GLA is then run on the two seed vectors until the distortion reduction value falls below threshold. Then each of the two resulting mean vectors are perturbed to yield four seed vectors for the next iteration of the splitting technique.
- the splitting technique is iterated until the desired number, in this case 1024, of codebook vectors is attained.
- the distortion and proximity measures used in step 14 can be perceptually weighted. For example, lower spatial frequency terms can be given more weight than higher spatial frequency terms. In addition, since this is vector rather than scalar quantization, interactive effects between the spatial frequency dimensions can be taken into account. Unweighted measures can be used if the transform space is perceptually linear, if no perceptual profile is available, or the decompressed data is to subject to further numeric processing before the image is presented for human viewing.
- the codebook designed in step 14 comprises a set of 1024 2x1 codebook vectors in the spatial frequency domain. These are arbitrarily assigned respective ten-bit indices at step 15. This completes codebook design procedure 10 of method M 1 for stage S 1.
- Fill-in procedure 20 for stage SI begins with step 21 of generating each distinct address to permit its contents to be determined.
- values are input into each of the tables in pairs. In alternative embodiments, some tables or all tables can have more inputs.
- the number of addresses is the product of the number of possible distinct values that can be received at each input. Typically, the number of possible distinct values is a power of two.
- Each input Vj is a scalar value corresponding to an intensity assigned to a respective pixel of an image.
- These inputs are concatenated at step 24 in pairs to define a two-dimensional vector (VJ, V(J+1)) in a spatial pixel domain. (Steps 22 and 23 are bypassed for the design of first-stage table Tl .)
- the input vectors must be expressed in the same domain as the codebook vectors, i.e., a two-dimensional spatial frequency domain. Accordingly, a DCT is applied at step 25 to yield a two-dimensional vector in the spatial frequency domain of the table Tl codebook.
- the table Tl codebook vector closest to this input vector is determined at step 26.
- the proximity measure is unweighted mean square error. Better performance is achieved using an objective measure like unweighted mean square error as the proximity measure during table building rather than a perceptually weighted measure.
- an unweighted proximity measurement is not required in general for this step.
- the measurement using during table fill at step 26 is weighted less on the average than the measures used in step 14 for codebook design.
- the index Wj assigned to the closest codebook vector at step 16 is then entered as the contents at the address corresponding to the input pair (Vj, V(j+1)). During operation of system Tl, it is this index that is output by table Tl in response to the given pair of input values.
- indexes Wj are assigned to all 65,536 addresses of table Tl, method Ml design of table Tl is complete.
- the codebook design begins with step 11 of selecting training images, just as for first-stage table Tl.
- the training images used for design of the table Tl codebook can be used also for the design of the second stage codebook.
- the training images are divided into 2x2 pixel blocks; the 2x2 pixel blocks are expressed as image vectors in four-dimensional vector space in a pixel domain; in other words, each of four vector values characterizes the intensity associated with a respective one of the four pixels of the 2x2 pixel block.
- the four-dimensional vectors are converted using a DCT to a spatial frequency domain.
- a four-dimensional pixel-domain vector can be expressed as a 2x2 array of pixels
- a four-dimensional spatial frequency domain vector can be expressed as a 2x2 array of spatial frequency functions: F00 FOI
- the four values of the spatial frequency domain respectively represent: F00)— an average intensity for the 2x2 pixel block; F01) ⁇ an intensity difference between the left and right halves of the block; F10)— an intensity difference between the top and bottom halves of the block; and Fl 1)— a diagonal intensity difference.
- the DCT conversion is lossless (except for small rounding errors) in that the spatial pixel domain can be retrieved by applying an inverse DCT to the spatial frequency domain vector.
- the four-dimensional frequency-domain vectors serve as the training sequence for second stage codebook design by the LBG/GLA algorithm.
- the proximity and distortion measures can be the same as those used for design of the codebook for table Tl .
- the difference is that for table T2, the measurements are performed in a four-dimensional space instead of a two-dimensional space.
- Eight-bit indices Xj are assigned to the codebook vectors at step 15, completing codebook design procedure 10 of method Ml.
- the address entries are to be determined using a proximity measure in the space in which the table T2 codebook is defined.
- the table T2 codebook is defined in a four-dimensional spatial frequency domain space.
- the address inputs to table T2 are pairs of indices (Wj,W(J+l)) for which no meaningful metric can be applied. Each of these indices corresponds to a table Tl codebook vector. Decoding indices (Wj,W(J+l)) at step 22 yields the respective table Tl codebook vectors, which are defined in a metric space.
- the table Tl codebook vectors are defined in a two-dimensional space, whereas four-dimensional vectors are required by step 26 for stage S2. While two two-dimensional vectors frequency domain can be concatenated to yield a four-dimensional vector, the result is not meaningful in the present context: the result would have two values corresponding to average intensities, and two values corresponding to left-right difference intensities; as indicated above, what would be required is a single average intensity value, a single left-right difference value, a single top-bottom difference value, and a single diagonal difference value.
- an inverse DCT is applied at step 23 to each of the pair of two-dimensional table Tl codebook vectors yielded at step 22.
- the inverse DCT yields a pair of two-dimensional pixel-domain vectors that can be meaningfully concatenated to yield a four-dimensional vector in the spatial pixel domain representing a 2x2 pixel block.
- a DCT transform can be applied, at step 25, to this four-dimensional pixel domain vector to yield a four-dimensional spatial frequency domain vector.
- This four-dimensional spatial frequency domain vector is in the same space as the table T2 codebook vectors. Accordingly, a proximity measure can be meaningfully applied at step 26 to determine the closest table T2 codebook vector.
- the index Xj assigned at step 15 to the closest table T2 codebook vector is assigned at step 27 to the address under consideration.
- table design method Ml for table T2 is complete.
- Table design method Ml for intermediate stage S3 is similar to that for intermediate stage S2, except that the dimensionality is doubled.
- Codebook design procedure 20 can begin with the selection of the same or similar training images at step 11.
- the images are converted to eight-dimensional pixel-domain vectors, each representing a 4x2 pixel block of a training image.
- a DCT is applied at step 13 to the eight-dimensional pixel-domain vector to yield an eight- dimensional spatial frequency domain vector.
- the array representation of this vector is:
- basis functions F00, FOI , F10, and Fl l have roughly, the same meanings as they do for a 2x2 array, once the array size exceeds 2x2, it is no longer adequate to describe the basis functions in terms of differences alone. Instead, the terms express different spatial frequencies.
- the functions, F00, FOI, F02, F03, in the first row represent increasingly greater horizontal spatial frequencies.
- the functions F00, FOI, in the first column represent increasingly greater vertical spatial frequencies.
- the remaining functions can be characterized as representing two-dimensional spatial frequencies that are products of horizontal and vertical spatial frequencies.
- a perceptual proximity measure might assign a relatively low (less than unity) weight to high spatial frequency terms such as F03 and F04.
- a relatively high (greater than unity) weight can be assigned to low spatial frequency terms.
- the perceptual weighting is used in the proximity and distortion measures during codebook assignment in step 14. Again, the splitting variation of the GLA is used.
- Table fill-in procedure 20 for table T3 is similar to that for table T2.
- Each address generated at step 21 corresponds to a pair (XJ, X(J+1)) of indices. These are decoded at step 22 to yield a pair of four-dimensional table T2 spatial-frequency domain codebook vectors at step 22.
- An inverse DCT is applied to these two vectors to yield a pair of four-dimensional pixel-domain vectors at step 23.
- the pixel domain vectors represent 2x2 pixel blocks which are concatenated at step 24 so that the resulting eight-dimensional vector in the pixel domain corresponds to a 4x2 pixel block.
- a DCT is applied to the eight-dimensional pixel domain vector to yield an eight-dimensional spatial frequency domain vector in the same space as the table T3 codebook vectors.
- the closest table T3 codebook vector is determined at step 26, preferably using an unweighted proximity measure such as mean-square error.
- the table T3 index Yj assigned at step 15 to the closest table T3 codebook vector is entered at the address under consideration at step 27. Once corresponding entries are made for all table T3 addresses, design of table T3 is complete.
- Table design method Ml for final-stage table T4 can begin with the same or a similar set of training images at step 11.
- the training images are expressed, at step 12, as a sequence of sixteen-dimensional pixel-domain vectors representing 4x4 pixel blocks (having the form of Bi in FIG. 1).
- a DCT is applied at step 13 to the pixel domain vectors to yield respective sixteen- dimensional spatial frequency domain vectors, the statistical profile of which is used to build the final-stage table T4 codebook.
- step 16 builds a tree-structured codebook.
- the main difference between tree-structured codebook design and the full-search codebook design used for the preliminary stages is that most of the codebook vectors are determined using only a respective subset of the training vectors.
- the mean, indicated at A in FIG. 3, of the training vectors is determined.
- the training vectors are in a sixteen-dimensional spatial frequency domain.
- the mean is perturbed to yield seed vectors for a two-vector codebook.
- the GLA is run to determine the codebook vectors for the two-vector codebook.
- the clustering of training vectors to the two-vector-codebook vectors is treated as permanent.
- Indices 0 and 1 are assigned respectively to the two-vector-codebook vectors, as shown in FIG. 3.
- Each of the two- vector-codebook vectors are perturbed to yield two pairs of seed vectors.
- the GLA is run using only the training vectors assigned to its parent codebook vector.
- the result is a pair of child vectors for each of the original two-vector-codebook vectors.
- the child vectors are assigned indices having as a prefix the index of the parent vector and a one bit suffice.
- the child vectors of the codebook vector assigned index 0 vector are assigned indices 00 and 01, while the child vectors of 1 codebook vector are assigned indices 10 and 11.
- the assignment of training vectors to the four child vectors is treated as permanent.
- the starting point for the pruning has the same general shape as the tree that results from the pruning.
- Such a tree can be obtained by the preferred "greedily-growing" variation, in which growth is node-by-node. In general, the growth is uneven, e.g., one sibling can have grandchildren before the other sibling has children.
- the determination of which childless node is the next to be grown involves computing a joint measure
- joint entropy and distortion measures are determined for two three-vector codebooks, each including an aunt and two nephews.
- One three-vector codebook includes vectors 0, 10, and 11; the other three-vector codebook includes vectors 1, 00, and 01.
- the three- vector codebook with the lower joint measure supersedes the two-vector codebook.
- the table T4 codebook is grown one vector at a time (instead of doubling each iteration as with the splitting procedure.)
- the parent that was replaced by her children is assigned an ordinal.
- the lower distortion is associated with the children of vector 1.
- the three vector codebook consists of vectors 11, 10, and 0.
- the ordinal 1 (in parenthesis in FIG. 3) is assigned to the replaced parent vector 1. This ordinal is used in selecting compression scaling.
- the two new codebook vectors e.g.,
- the GLA is run on each pair using only training vectors assigned to the respective parent.
- the result is two pairs of proposed new codebook vectors (111, 110) and (101 ,100).
- Distortion measures are obtained for each pair. These distortions measures are compared with the already obtained distortion measure for the vector, e.g., 0, common to the two-vector and three-vector codebooks.
- the tree is grown from the codebook vector for which the growth yields the least distortion. In the example of FIG. 3, the tree is grown from vector 0, which is assigned the ordinal 2.
- FIG. 3 shows a tree after nine iterations of the tree-growing procedure.
- tree growth can terminate with a tree with the desired number, of end nodes corresponding to codebook vectors is achieved.
- the resulting tree is typically not optimal.
- growth continues well past the size required for the desired codebook.
- the average bit length for codes associated with the overgrown three can be twice the average bit length desired for the tree to be used for the maximum precision code.
- the overgrown tree can be pruned node-by-node using a joint measure of distortion and entropy until a tree of the desired size is achieved. Note that the pruning can also be used to obtain an entropy shaped tree from an evenly overgrown tree.
- Lower precision trees can be designed by the ordinals assigned during greedy growing. There may be some gaps in the numbering sequence, but a numerical order is still present to guide selection of nodes for the lower-precision trees. Preferably, however, the high-precision tree is pruned using the joint measure of distortion and entropy to provide better low-precision trees. To the extent of the pruning, ordinals can be reassigned to reflect pruning order rather than the growing order. If the pruning is continued to the common ancestor and its children, then all ordinals can be reassigned according to pruning order.
- the full-precision-tree codebook provides lower distortion and a lower bit rate than any of its predecessor codebooks. If a higher bit rate is desired, one can select a suitable ordinal and prune all codebook vectors with higher ordinals.
- the resulting predecessor codebook provides a near optimal tradeoff of distortion and bit rate.
- a 1024-vector codebook is built, and its indices are used for index ZA.
- index ZB the tree is pruned back to ordinal 512 to yield a higher bit rate.
- ZC the index is pruned back to ordinal 256 to yield an even higher bit rate.
- the code pruner 51 of decoder DEC has information regarding the ordinals to allow it to make appropriate bit-rate versus distortion tradeoffs.
- indices ZA, ZB, and ZC could be entered in sections of respective addresses of table T4, doing so would not be memory efficient. Instead ZC, Zb, and Za are stored. Zb indicates the bits to be added to index ZC to obtain index ZB. Za indicates the bits to be added to index ZB to obtain index ZA.
- Fill-in procedure 20 for table T4 begins at step 21 with the generation of the 2 20 addresses corresponding to all possible distinct pairs of inputs (Y1 ,Y2).
- Each third stage index Yj is decoded at step 22 to yield the respective eight-dimensional spatial-frequency domain table T3 codebook vector.
- An inverse DCT is applied at step 23 to these table T3 codebook vectors to obtain the corresponding eight-dimensional pixel domain vectors representing 4x2 pixel blocks.
- These vectors are concatenated at step 24 to form a sixteen-dimensional pixel-domain vector corresponding to a respective 4x4 pixel block.
- a DCT is applied at step 24 to yield a respective sixteen-dimensional spatial frequency domain vector in the same space as the table T4 codebook.
- the closest table T4 codebook vector in each of the three sets of codebook vectors are identified at step 26, using an unweighted proximity measure.
- the class indices ZA, ZB, and AC associated with the closest codebook vectors are assigned to the table T4 address under consideration. Once this assignment is iterated for all table T4 addresses, design of table T4 is complete. Once all tables T1-T4 are complete, design of hierarchical table HLT is complete.
- VRTSHVQ variable-rate tree-structured hierarchical table-based vector quantization
- the tables used to implement vector quantization can also implement block transforms.
- table lookup encoders input vectors to the encoders are used directly as addresses in code tables to choose the codewords. There is no need to perform the forward or reverse transforms. They are implemented in the tables.
- Hierarchical tables can be used to preserve manageable table sizes for large dimension VQ's to quantize a vector in stages. Since both the encoder and decoder are implemented by table lookups, there are no arithmetic computations required in the final system implementation.
- the algorithms are a novel combination of any generic block transform (DCT, Haar, WHT) and hierarchical vector quantization. They use perceptual weighting and subjective distortion measures in the design of VQ's. They are unique in that both the encoder and the decoder are implemented with only table lookups and are amenable to efficient software and hardware solutions.
- VQ Full-search vector quantization
- a transform code is a structured vector quantizer in which the encoder performs a linear transformation followed by scalar quantization of the transform coefficients.
- This structure also increases the decoder complexity, however, since the decoder must now perform an inverse transform.
- transform coding the computational complexities of the encoder and decoder are essentially balanced, and hence transform coding finds natural application to point-to-point communication, such as video telephony.
- a special advantage of transform coding is that perceptual weighting, according to frequency sensitivity, is simple to perform by allocating bits appropriately among transform coefficients.
- a number of other structured vector quantization schemes decrease encoder complexity but do not simultaneously increase decoder complexity.
- Such schemes include tree-structured VQ, lattice VQ, fine-to-coarse VQ, etc.
- Hierarchical table-based vector quantization replaces the full-search encoder with a hierarchical arrangement of table lookups, resulting in a maximum of one table lookup per sample to encode. The result is a balanced scheme, but with extremely low computational complexity at both the encoder and decoder.
- the hierarchical arrangement allows efficient encoding for multiple rates.
- HVQ finds natural application to collaborative video over heterogeneous networks of inexpensive general pu ⁇ ose computers.
- Perceptually significant distortion measures can be integrated into HTBVQ based on weighting the coefficients of arbitrary transforms. Essentially, the transforms are pre-computed and built into the encoder and decoder lookup tables. Thus gained are the perceptual advantages of transform coding while maintaining the computational simplicity of table lookup encoding and decoding.
- HTBVQ is a method of encoding vectors using only table lookups.
- the r m - 1 -bit outputs from the previous stage are combined into blocks of length k ⁇ to directly address a lookup table with ⁇ , address bits to produce r m output bits per block.
- the r m bits output from the final stage M may be sent directly through the channel to the decoder, if the quantizer is a fixed-rate quantizer, or the bits may be used to index a table of variable-length codes, for example, if the quantizer is a variable- rate quantizer.
- the computational complexity of the encoder is at most one table lookup per input symbol
- the table at stage can be regarded as a mapping from two input indices iTM -1 and iTM '1 , each in ⁇ 0,1, ,255 ⁇ , to an output index i m also in ⁇ 0,1 , ,255 ⁇ .
- HTBVQ is ideally suited to implementing perceptually meaningful, if complex, distortion measures.
- d'(x, x) as a function of x has a Taylor series expansion around x in which the constant and first order terms are zero, and the quadratic term is non-negative semi-definite.
- W x (w,, w k ) and K is the dimension of x.
- T the transformation matrix of some fixed transform, such as the Haar, Walsh-Hadamard, or discrete cosine transform, and we shall let the weights W x vary arbitrarily with x. This is a reasonably general class of perceptual distortion measures.
- the weights reflect human visual sensitivity to quantization errors in different transform coefficients, or bands.
- the weights may be input-dependent to model masking effects.
- the weights control an effective stepsize, or bit allocation, for each band.
- stepsize or bit allocation, for each band.
- bits are allocated between bands in accordance with the strength of the signal in the band and an appropriate perceptual model.
- the weights w, ,W K play a role corresponding to the stepsizes.
- each encoding cell has the same volume V in K-space.
- V m times a sphere packing coefficient less than 1 in the scaled space.
- each encoding cell has roughly linear dimension -w°/ 5 V m along the jth coordinate.
- HTBVQ can be combined with block based transforms like the DCT, the Haar and the
- Walsh-Hadamard Transform perceptually weighted to improve visual performance.
- WTHVQ Weighted Transform HVQ
- the encoder of a WTHVQ consists of M stages (as in FIG. 1), each stage being implemented by a lookup table. For image coding, separable transforms are employed, so the odd stages operate on the rows while the even stages operate on the columns of the image.
- the first stage gives a compression of 2: 1.
- the second stage corresponds to a 2x2 transform on the input image followed by perceptually weighted vector quantization using a subjective distortion measure, with 256 codewords. The only difference is that the 2x2 vector is quantized successively in two stages.
- the compression achieved after the second stage is 4: 1.
- stage i 1 ⁇ i ⁇ M
- Stage i corresponds to a 2 , 2 x2 l 2 perceptually weighted transform, for i even, or a 2 (,+I) 2 ⁇ 2° "I)/2 transform, for i odd, followed by a perceptually weighted vector quantizer using a subjective distortion measure with 256 codewords. The only difference is that the quantization is performed successively in i stages. The compression achieved after stage i is 2': 1.
- the last stage produces the encoding index u, which represents an approximation to the input (perceptually weighted transform) vector and sends it to the decoder.
- This encoding index is similar to that obtained in a direct transform VQ with an input weighted distortion measure.
- the decoder of a WTHVQ is the same as a decoder of such a transform VQ. That is, it is a lookup table in which the reverse transform is done ahead of time on the codewords.
- the computational and storage requirements of WTHVQ are same as that of ordinary
- the design of a WTHVQ consists of two major steps.
- the first step designs VQ codebooks for each transform stage. Since each perceptually weighted transform VQ stage has a different dimension and rate they are designed separately. A subjectively meaningful distortion measure as described above is used for designing the codebooks.
- the codebooks for each stage of the WTHVQ are designed independently by the generalized Lloyd algorithm (GLA) run on the transform of the appropriate order on the training sequence.
- the first stage codebook with 256 codewords is designed by running GLA on a 2x1 transform (DCT, Haar, or WHT) of the training sequence.
- the stage i codebook (256 codewords) is designed using the GLA on a transform of the training sequence of the appropriate order for that stage.
- the reconstructed codewords for the transformed data using the subjective distortion measure - are given by:
- the original training sequence is used to design all stages by transforming it using the corresponding transforms of the appropriate order for each stage.
- the corresponding input training sequence to each stage are generally different because each stage has to go through a lot of previous stages and the sequence is quantized successively in each stage and is hence different at each stage.
- the second step in the design of WTHVQ builds lookup tables from the designed codebooks. After having built each codebook for the transform the corresponding code tables are built for each stage.
- the first stage table is built by taking different combinations of two 8-bit input pixels. There are 2 16 such combinations. For each combination a 2x1 transform is performed. The index of the codeword closest to the transform for the combination in the sense of minimum distortion rule (subjective distortion measure d,.) is put in the output entry of the table for that particular input combination. This procedure is repeated for all possible input combinations.
- Each output entry (2 16 total entries) of the first stage table has 8 bits.
- the second stage table operates on the columns.
- the product combination of two first stage tables is taken by taking the product of two 8-bit outputs from the first stage table.
- a successively quantized 2x2 transform is obtained by doing a 2x1 inverse transform on the two codewords obtained by using the indices for the first stage codebook.
- Now on the 2x2 raw data obtained a 2x2 transform is performed and the index of the codeword closest to this transformed vector in the sense of the subjective distortion measure d ⁇ is put in the corresponding output entry. This procedure is repeated for all input entries in the table.
- Each output entry for the second stage table also has 8 bits.
- the third stage table operates on the rows.
- the product combination of two second stage tables is obtained by taking the product of the output entries of the second stage tables.
- Each output entry of the second stage table has 8 bits.
- the total number of different input entries to the third stage table are 2 16 .
- a successively quantized 4x2 transform is obtained by doing a 2x2 inverse transform on the two codewords obtained by using the indices for the second stage codebook. Now on the 4x2 raw data obtained a 4x2 transform is performed and the index of the codeword closest in the sense of the subjective distortion measure d to this transformed vector is put in the corresponding output entry.
- All remaining stage tables are built in a similar fashion by performing two inverse transforms and then performing a forward transform on the data.
- the nearest codeword to this transform data in the sense of subjective distortion measure d-- is obtained from the codebook for that stage and the corresponding index is put in the table.
- the last stage table has the index of the codeword as its output entry which is sent to the decoder.
- the decoder has a copy of the last stage codebook and uses the index for the last stage to output the corresponding codeword.
- a simpler table building procedure can be used for the Haar and the Walsh-Hadamard transforms. This happens because of the nice property of the Haar and WHT that higher order transform can be obtained as a linear combination of a lower order transform on the partitioned data.
- the table building for the DCT i.e. the inverse transform method, will be more expensive than the Haar and the WHT because at each stage two inverse transforms and one forward DCT transform must be performed.
- VQ are the same at each compression ratio. This is because the transforms are all orthogonal, any differences are due to the fact that the splitting algorithm in the GLA is sensitive to the coordinate system. JPEG performs around 5 dB better than these schemes since it is a variable rate code. These VQ based algorithms being fixed rate have other advantages compared to JPEG. However by using entropy coding along with these algorithms 25% more compression can be achieved.
- Table III gives the PSNR results on Lena for different compression ratios for plain HVQ, unweighted Haar VQ, unweighted WHT HVQ and unweighted DCT HVQ. It can be seen from Table HI that the PSNR results of transform HVQ are the same as the plain HVQ results for the same compression ratio. Comparing the results of Table III with Table II we find that the HVQ based schemes perform around 0.7 dB worse than the full search VQ schemes.
- Table IV gives the PSNR results on Lena for different compression ratios for full search plain VQ, perceptually weighted full search Haar VQ, perceptually weighted full-search WHT VQ and perceptually weighted full search DCT VQ.
- the weighting increases the subjective quality of the compressed images, though it reduces the PSNR.
- the subjective quality of the images compressed using weighted VQ's is much better than the unweighted VQ's.
- Table IV also gives the PSNR results on Lena for different compression ratios for perceptually weighted Haar VQ, WHT HVQ and DCT HVQ.
- the visual quality of the compressed images obtained using weighted transform HVQ's is significantly higher than for plain HVQ.
- the quality of the weighted transform VQ's compressed images is about the same as that of the weighted transform HVQ's compressed images.
- Table V gives the encoding times of the different algorithms on a SUN Sparc- 10 workstation on Lena. It can be seen from Table V that the encoding times of the transform HVQ and plain HVQ are same. It takes 12 ms for the first stage encoding, 24 ms for the second stage encoding and so on. On the other hand JPEG requires 250 ms for encoding at all compression ratios. Thus the HVQ based encoders are 10-25 times faster than a JPEG encoder. The HVQ based encoders are also around 50-100 times faster than full search VQ based encoders. This low computational complexity of HVQ is very useful for collaborative video over heterogeneous networks. It makes 30 frames per second software only video encoding possible on general pu ⁇ ose workstations.
- Table VI gives the decoding times of different algorithms on a SUN Sparc- 10 workstation on Lena. It can be seen from Table VI that the decoding times of the transform HVQ, plain HVQ, plain VQ and transform VQ are same. It takes 13 ms for decoding a 2: 1 compressed image, 16 ms for decoding a 4:1 compressed image and so on. On the other hand JPEG requires 200 ms for decoding at all compression ratios. Thus the HVQ based decoders are 20-40 times faster than a JPEG decoder. The decoding times of transform VQ are same as that of plain VQ as the transforms can be precomputed in the decoder tables. This low computational complexity of HVQ decoding again allows 30 frames per second video decoding in software.
- VQ product VQ
- mean-removed VQ multi-stage VQ
- hierarchical VQ non-linear inte ⁇ olative
- VQ predictive VQ
- weighted universal VQ weighted universal VQ.
- entropy-constrained VQ to get a variable rate code
- tree- structured VQ to get an embedded code
- classified VQ, product VQ, mean-removed VQ, multi-stage VQ, hierarchical VQ and non-linear inte ⁇ olative VQ are considered to overcome the complexity problems of unconstrained VQ and thereby allow the use of higher vector dimensions and larger codebook sizes.
- Recursive vector quantizers such as predictive VQ achieve the performance of a memory-less VQ with a large codebook while using a much smaller codebook.
- Weighted universal VQ provide for multi-codebook systems.
- Perceptually weighted hierarchical table-lookup VQ can be combined with different con ⁇ strained and recursive VQ structures.
- the HVQ encoder still consists of M stages of table lookups. The last stage differs for the different forms of VQ structures.
- Entropy-constrained vector quantization (ECVQ) , which minimizes the average distortion subject to a constraint on the entropy of the codewords, can be used to obtain a variable- rate system.
- ECHVQ has the same structure as HVQ, except that the last stage codebook and table are variable-rate.
- the last stage codebook and table are designed using the ECVQ algorithm, in which an unconstrained minimization problem is solved: min(D+ ⁇ H), where D is the average distortion (obtained by taking expected value of d defined above and H is the entropy.
- min(D+ ⁇ H) an unconstrained minimization problem is solved: min(D+ ⁇ H)
- D the average distortion (obtained by taking expected value of d defined above
- H is the entropy.
- this modified distortion measure is used in the design of the last stage codebook and table.
- the last stage table outputs a variable length index which is sent to the decoder.
- the decoder has a copy of the last stage codebook and uses the index for the last stage to output the corresponding codeword.
- the design of an ECHVQ consists of two major steps. The first step designs VQ codebooks for each stage. Since each VQ stage has a different dimension and rate they are designed separately. As described above, a subjectively meaningful distortion measure is used for designing the codebooks. The codebooks for each stage except the last stage of the ECHVQ are designed independently by the generalized Lloyd algorithm (GLA) run on the appropriate vector size of the training sequence. The last stage codebook is designed using the ECVQ algorithm.
- the second step in the design of ECHVQ builds lookup tables from the designed codebooks. After having built each codebook the corresponding code tables are built for each stage. All tables except the last stage table are built using the procedure described above. The last stage table is designed using a modified distortion measure. In general the last stage table implements the mapping
- r M (i) is the number of bits representing the i lh codeword in the last stage codebook. Only the last stage codebook and table need differ for different values of lambda.
- a tree-structured VQ at the last stage of HVQ can be used to obtain an embedded code.
- the codewords lie in an unstructured codebook, and each input vector is mapped to the minimum distortion codeword. This induces a partition of the input space into Voronoi encoding regions.
- TSVQ on the other hand, the codewords are arranged in a tree structure, and each input vector is successively mapped (from the root node) to the minimum distortion child node. This induces a hierarchical partition, or refinement of the input space as the depth of the tree increases.
- an input vector mapping to a leaf node can be represented with high precision by the path map from the root to the leaf, or with lower precision by any prefix of the path.
- TSVQ produces an embedded encoding of the data. If the depth of the tree is R and the vector dimension is k, then bit rates 0/k, 1 k, , R/k, can all be achieved.
- Variable-rate TSVQs can be constructed by varying the depth of the tree. This can be done by "greedily growing" the tree one node at a time (GGTSVQ), or by growing a large tree and pruning back to minimize its average distortion subject to a constraint on its average length (PTSVQ) or entropy (EPTSVQ).
- the last stage table outputs a fixed or variable length embedded index which is sent to the decoder.
- the decoder has a copy of the last stage tree-structured codebook and uses the index for the last stage to output the corresponding codeword.
- TSHVQ has the same structure as HVQ except that the last stage codebook and table are tree-structured.
- the last stage table outputs a fixed or variable length embedded index which is transmitted on the channel.
- the design of a TSHVQ again consists of two major steps. The first step designs VQ codebooks for each stage. The codebooks for each stage except the last stage of the TSHVQ are designed independently by the generalized Lloyd algorithm (GLA) run on the appropriate vector size of the training sequence. The second step in the design of TSHVQ builds lookup tables from the designed codebooks. After having built each codebook, the corresponding code tables are built for each stage. All tables except the last stage table are built using the procedure described above.
- GLA generalized Lloyd algorithm
- the last stage table is designed by setting i M (i, M ⁇ ',i 2 u' ') to the variable length index i to which the concatenated vector ⁇ (i TM '1 ), ⁇ . ⁇ ' ) is encoded by the tree structured codebook.
- a classifier In Classified Hierarchical Table-Lookup VQ (CHVQ), a classifier is used to decide the class to which each input vector belongs. Each class has a set of HVQ tables designed based on codebooks for that class.
- the classifier can be a nearest neighbor classifier designed by GLA or an ad hoc edge classifier or any other type of classifier based on features of the vector, e.g. , mean and variance.
- the CHVQ encoder decides which class to use and sends the index for the class as side information.
- the advantage of classified VQ has been in reducing the encoding complexity of full-search VQ by using a smaller codebook for each class.
- CHVQ bit allocation can be done to decide the rate for a class based on the semantic significance of that class.
- the encoder sends side-information to the decoder about the class for the input vector.
- the class determines which hierarchy of tables to use.
- the last stage table outputs a fixed or variable length index which is sent to the decoder.
- the decoder has a copy of the last stage codebook for the different classes and uses the index for the last stage to output the corresponding codeword from the class codebook based on the received classification information.
- CHVQ has the same structure as HVQ except that each class has a separate set of HVQ tables.
- the last stage table outputs a fixed or variable (entropy-constrained CHVQ) length index which is sent to the decoder.
- the design of a CHVQ again consists of two major steps. The first step designs VQ codebooks for each stage for each class as for HVQ or ECHVQ. After having built each codebook the corresponding code tables are built for each stage for each class as in HVQ or ECHVQ.
- Product Hierarchical Table Lookup VQ reduces the storage complexity in coding a high dimensional vector by splitting the vector into two or more components and encode each split vector independently.
- an 8x8 block can be encoded as four 4x4 blocks, each encoded using the same set of HVQ tables for a 4x4 block.
- the input vector can be split into sub- vectors of varying dimension where each sub- vector will be encoded using the HVQ tables to the appropriate stage.
- the table and codebook design in this case is exactly the same as for HVQ.
- Mean-Removed Hierarchical Table-Lookup VQ (MRHVQ) is a form of product code to reduce the encoding and decoding complexity. It allows coding higher dimensional vectors at higher rates.
- MRHVQ is a mean-removed VQ in which the full search encoder is replaced by table-lookups.
- the first stage table outputs an 8-bit index for a residual and an 8-bit mean for a 2x1 block.
- the 8-bit index for the residual is used to index the second stage table.
- the output of the second stage table is used as input to the third stage.
- the 8-bit means for several 2x1 blocks after the first stage are further averaged and quantized for the input block and transmitted to the decoder independently of the residual index.
- the last stage table outputs a fixed or variable length (entropy-constrained MRHVQ) residual index which is sent to the decoder.
- the decoder has a copy of the last stage codebook and uses the index for the last stage to output the corresponding codeword from the codebook and adds the received mean of the block.
- MRHVQ has the same structure as the HVQ except that all codebooks and tables are designed for mean-removed vectors.
- the design of a MRHVQ again consists of two major steps. The first step designs VQ codebooks for each stage as for HVQ or ECHVQ on the mean- removed training set of the appropriate dimension. After having built each codebook the corresponding code tables are built for each stage as in HVQ or ECHVQ.
- Multi-Stage Hierarchical Table-Lookup VQ (MSHVQ) is a form of product code which allows coding higher dimensional vectors at higher rates.
- MSHVQ is a multi-stage VQ in which the full search encoder is replaced by a table-lookup encoder.
- the encoding is performed in several stages. In the first stage the input vector is coarsely quantized using a set of
- the first stage index is transmitted as coarse-level information.
- the residual between the input and the first stage quantized vector is again quantized using another set of HVQ tables. Note that the residual can be obtained through table-lookups at the second stage).
- the second stage index is sent as refinement information to the decoder. This procedure continues in which the residual between successive stages is encoded using a new set of HVQ tables. There is a need for bit-allocation between the different stages of MSHVQ.
- the decoder uses the transmitted indices to look up the corresponding codebooks and adds the reconstructed vectors.
- MSHVQ has the same structure as the HVQ except that it has several stages of HVQ.
- each stage outputs a fixed or variable (entropy-constrained MSHVQ) length index which is sent to the decoder.
- the design of a MSHVQ consists of two major steps.
- the first stage encoder codebooks are designed as in HVQ.
- the second stage codebooks are designed closed loop by using the residual between the training set and the quantized training set after the first stage. After having built each codebook the corresponding code tables are built for each stage essentially as in HVQ or ECHVQ. The only difference is that the tables for the second and subsequent stages are designed for residual vectors.
- H-HVQ Hierarchical-Hierarchical Table-Lookup VQ
- H-HVQ is a hierarchical VQ in which the full search encoder is replaced by a table-lookup encoder.
- MSHVQ the H-HVQ encoding is performed in several stages.
- a large input vector (super-vector) is coarsely quantized using a set of HVQ tables to give a quantized feature vector.
- the first stage index is transmitted to the decoder.
- the residual between the input and the first stage quantized vector is again quantized using another set of HVQ tables but the super-vector is split into smaller sub- vectors.
- the residual can be obtained through table-lookups at the second stage.
- the second stage index is also sent to the decoder. This procedure of partitioning and quantizing the super-vector by encoding the successive residuals is repeated for each stage. There is a need for bit-allocation between the different stages of H-HVQ.
- the decoder uses the transmitted indices to look up the corresponding codebooks and adds the reconstructed vectors.
- the structure of H- HVQ encoder is similar to that of MSHVQ except that in this case the vector dimensions at the first stage and subsequent stages of encoding differ.
- the design of a H-HVQ is same as that of MSHVQ with the only difference is that the vector dimension reduces in subsequent stages.
- Non-linear Interpolative Table-Lookup VQ allows a reduction in encoding and storage complexity compared to HVQ.
- NLHVQ is a non-linear inte ⁇ olative VQ in which the full- search encoder is replaced by a table-lookup encoder.
- the encoding is performed as in HVQ, except that a feature vector is extracted from the original input vector and the encoding is performed on the reduced dimension feature vector.
- the last stage table outputs a fixed or variable length (entropy-constrained NIHVQ) index which is sent to the decoder.
- the decoder has a copy of the last stage codebook and uses the index for the last stage to output the corresponding codeword.
- the decoder codebook has the optimal non-linear inte ⁇ olated codewords of the dimension of the input vector.
- the design of a NIHVQ consists of two major steps.
- the first step designs encoder VQ codebooks from the feature vector for each stage as for HVQ or ECHVQ.
- the last stage codebook is designed using nonlinear inte ⁇ olative VQ. After having built each codebook the corresponding code tables are built for each stage for each class as in HVQ or ECHVQ.
- Predictive Hierarchical Table-Lookup VQ is a VQ with memory.
- the only difference between PHVQ and predictive VQ (PVQ) is that the full search encoder is replaced by a hierarchical arrangement of table-lookups.
- PHVQ takes advantage of the inter-block correlation in images.
- PHVQ achieves the performance of a memory-less VQ with a large codebook while using a much smaller codebook.
- the current block is predicted based on the previously quantized neighboring blocks using linear prediction and the residual between the current block and its prediction is coded using HVQ.
- the prediction can also performed using table-lookups and the quantized predicted block is used for calculating the residual again through table-lookups.
- the last stage table outputs a fixed or variable length index for the residual which is sent to the decoder.
- the decoder has a copy of the last stage codebook and uses the index for the last stage to output the corresponding codeword from the codebook.
- the decoder also predicts the current block from the neighboring blocks using table-lookups and adds the received residual to the predicted block.
- PHVQ all codebooks and tables are designed for the residual vectors.
- the last stage table outputs a fixed or variable (entropy-constrained PHVQ) length index which is sent to the decoder.
- the design of a PHVQ consists of two major steps. The first step designs VQ codebooks for each stage as for HVQ or ECHVQ on the residual training set of the appropriate dimension (closed-loop codebook design). After having built each codebook the corresponding code tables are built for each stage as in HVQ or ECHVQ, the only difference is that the residual can be calculated in the first stage table.
- Weighted Universal Hierarchical Table-Lookup VQ is a multiple-codebook VQ system in which a super- vector is encoded using a set of HVQ tables and the one which minimize the distortion is chosen to encode all vectors within the super-vector. Side-information is sent to inform the decoder about which codebook to use.
- WUHVQ is a weighted universal VQ (WUVQ) in which the selection of codebook for each super- vector and the encoding of each vector within the super-vector is done through table-lookups.
- the last stage table outputs a fixed or variable length (entropy-constrained WUHVQ) index which is sent to the decoder.
- the decoder has a copy of the last stage codebook for the different tables and uses the index for the last stage to output the corresponding codeword from the selected codebook based on the received side- information.
- WUHVQ has multiple sets of HVQ tables.
- the design of a WUHVQ again consists of two major steps.
- the first step designs WUVQ codebooks for each stage as for HVQ or ECHVQ.
- After having built each codebook the corresponding HVQ tables are built for each stage for each set of HVQ tables as in HVQ or ECHVQ.
- FIGS. 4-8 show the PSNR (peak signal-noise-ratio) results on the 8-bit monochrome image Lena (512x512) as a function of bit- rate for the different algorithms.
- the codebooks for the VQs have been generated by training on 10 different images. PSNR results are given for unweighted VQs; weighting reduces the PSNR though the subjective quality of compressed images improves significantly. One should however note that there is about 2 dB equivalent gain in PSNR by using a subjective distortion measure.
- FIG. 4 gives the PSNR results on Lena for greedily-grown-then pruned, variable-rate, tree-structured hierarchical vector quantization (VRTSHVQ). The results are for 4x4 blocks where the last stage is tree-structured. VRTSHVQ gives an embedded code at the last stage.
- VRTSHVQ again gains over HVQ. There is again about 0.5-0.7 dB loss compared to non- hierarchical variable-rate tree-structured table-based vector quantization (VRTSVQ).
- FIG. 5 gives the PSNR results on Lena for different bit-rates for plain VQ and plain HVQ.
- the results are on 4x4 blocks. We find that the HVQ performs around 0.5-0.7 dB worse than the full search VQ.
- FIG. 4 also gives the PSNR results on Lena for entropy-constrained HVQ (ECHVQ) with 256 codewords at the last stage. The results are on 4x4 blocks where the first three stages of ECHVQ are fixed-rate and the last stage is variable rate. It can be seen that ECHVQ gains around 1.5 dB over HVQ. There is however again a 0.5-0.7 dB loss compared to ECVQ.
- ECHVQ entropy-constrained HVQ
- Classified HVQ performs slightly worse than HVQ in rate-distortion but has the advantage of lower complexity (encoding and storage) by using smaller codebooks for each class.
- Product HVQ again performs worse in rate-distortion complexity compared to HVQ but has much lower encoding and storage complexity compared to HVQ as it partitions the input vector into smaller sub- vectors and encodes each one of them using a smaller set of HVQ tables.
- Mean-removed HVQ (MRHVQ) again performs worse in rate-distortion compared to HVQ but allows coding higher dimensional vectors at higher rates using the HVQ structure.
- FIG. 6 gives the PSNR results on Lena for hierarchical-HVQ (H-HVQ).
- the results are for 2-stage H-HVQ.
- the first stage operates on 8x8 blocks and is coded using HVQ to 8 bits.
- the residual is coded again using another set of HVQ tables.
- Figure 1 1 shows the results at different stages of the second-stage H-HVQ (each stage is coded to 8 bits).
- Fixed- rate H-HVQ gains around 0.5-1 dB over fixed- rate HVQ at most rates.
- Multi-stage HVQ (MSHVQ) is identical to H-HVQ where the second stage is coded to the original block size.
- FIG. 11 There is again about 0.5-0.7 dB loss compared to full search Shoham-Gersho HVQ results.
- FIG. 7 gives the PSNR results on Lena for entropy-constrained predictive HVQ
- FIG. 8 gives the PSNR results for entropy-constrained weighted-universal HVQ
- ECWUHVQ The super-vector is 16x16 blocks for these simulations and the smaller blocks are 4x4. There are 64 codebooks each with 256 4x4 codewords. It can be seen that ECWUHVQ gains around 3 dB over fixed-rate HVQ and 1.5 dB over ECHVQ. There is however again a 0.5- 0.7 dB loss compared to WUVQ.
- the encoding times of the transform HVQ and plain HVQ are same. It takes 12 ms for the first stage encoding, 24 ms for the first two stages and 30 ms for the first four stages of encoding a 512x512 image on a Sparc- 10 Workstation. On the other hand JPEG requires 250 ms for encoding at similar compression ratios.
- the encoding complexity of constrained and recursive HVQs increases by a factor of 2-8 compared to plain HVQ.
- the HVQ based encoders are around 50-100 times faster than their corresponding full search VQ encoders.
- the decoding times of the transform HVQ, plain HVQ, plain VQ and transform VQ are same. It takes 13 ms for decoding a 2:1 compressed image, 16 ms for decoding a 4:1 compressed image and 6 ms for decoding a 16:1 compressed 512x512 image on a Sparc- 10 Workstation.
- JPEG requires 200 ms for decoding at similar compression ratios.
- the decoding complexity of constrained and recursive HVQs does not increase much compared to that of HVQ.
- the HVQ based decoders are around 20-30 times faster than a JPEG decoder.
- the decoding times of transform VQs are same as that of plain VQs as the transforms can be precomputed in the decoder tables.
- constrained and recursive HVQ structures overcome the problems of fixed-rate memory-less VQ.
- the main advantage of these algorithms is very low computational complexity compared to the corresponding VQ structures.
- Entropy-constrained HVQ gives a variable rate code and performs better than HVQ.
- Tree-structured HVQ gives an embedded code and performs better than HVQ.
- Classified HVQ, product HVQ, mean-removed HVQ, multi-stage HVQ, hierarchical HVQ and non- linear inte ⁇ olative HVQ overcome the complexity problems of unconstrained VQ and allow the use of higher vector dimensions and achieve higher rates.
- Predictive HVQ achieves the performance of a memory-less VQ with a large codebook while using a much smaller codebook. It provides better rate-distortion performance by taking advantage of inter- vector correlation. Weighted universal HVQ again gains significantly over HVQ in rate- distortion. Further some of these algorithms (e.g. PHVQ, WUHVQ) with subjective distortion measures perform better or comparable to JPEG in rate-distortion at a lower decoding complexity.
- constrained and recursive vector quantizer encoders implemented by table-lookups.
- These vector quantizers include entropy constrained VQ, tree-structured VQ, classified VQ, product VQ, mean-removed VQ, multi-stage VQ, hierarchical VQ, non-linear interpolative VQ, predictive VQ and weighted-universal VQ.
- Our algorithms combine these different VQ structures with hierarchical table-lookup vector quantization. This combination significantly reduces the complexity of the original VQ structures.
- the transforms are pre-computed and built into the encoder and decoder lookup tables.
Abstract
Description
Claims
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP97917006A EP0890222A1 (en) | 1996-03-28 | 1997-03-26 | Table-based compression with embedded coding |
AU25472/97A AU2547297A (en) | 1996-03-28 | 1997-03-26 | Table-based compression with embedded coding |
JP9534591A JP2000507754A (en) | 1996-03-28 | 1997-03-26 | Table-based compression with embedded coding |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US08/623,299 | 1996-03-28 | ||
US08/623,299 US6215910B1 (en) | 1996-03-28 | 1996-03-28 | Table-based compression with embedded coding |
Publications (1)
Publication Number | Publication Date |
---|---|
WO1997036376A1 true WO1997036376A1 (en) | 1997-10-02 |
Family
ID=24497537
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/US1997/004879 WO1997036376A1 (en) | 1996-03-28 | 1997-03-26 | Table-based compression with embedded coding |
Country Status (5)
Country | Link |
---|---|
US (6) | US6215910B1 (en) |
EP (1) | EP0890222A1 (en) |
JP (1) | JP2000507754A (en) |
AU (1) | AU2547297A (en) |
WO (1) | WO1997036376A1 (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO1999027715A1 (en) * | 1997-11-21 | 1999-06-03 | Sharp Laboratories Of America, Inc. | Method and apparatus for compressing reference frames in an interframe video codec |
EP0933946A2 (en) * | 1998-01-29 | 1999-08-04 | Xerox Corporation | A method for transmitting data using an embedded bit stream produced in a hierarchical tablelookup vector quantizer |
EP1009167A2 (en) * | 1998-12-07 | 2000-06-14 | Xerox Corporation | Method and apparatus for electronic data compression |
WO2001031497A1 (en) * | 1999-10-22 | 2001-05-03 | Activesky, Inc. | An object oriented video system |
US6272180B1 (en) | 1997-11-21 | 2001-08-07 | Sharp Laboratories Of America, Inc. | Compression and decompression of reference frames in a video decoder |
WO2002075929A1 (en) * | 2001-03-16 | 2002-09-26 | Coression Oy | A method for compressing data |
WO2006024077A2 (en) * | 2004-08-31 | 2006-03-09 | Media World Communications Limited | System and method for encoding and decoding video |
JP2010181889A (en) * | 1998-08-24 | 2010-08-19 | Mindspeed Technologies Inc | Selection of scalar quantization (sq) and vector quantization (vq) for speech encoding |
CN102496368A (en) * | 2011-12-20 | 2012-06-13 | 重庆金美通信有限责任公司 | Improved vector quantization method |
US8219551B2 (en) | 2007-10-31 | 2012-07-10 | General Instrument Corporation | Decoding a hierarchical multi-layer data package |
CN102740325A (en) * | 2011-04-02 | 2012-10-17 | 华为技术有限公司 | Method, device for acquiring channel information and method, device for optimizing beam forming |
Families Citing this family (159)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6965644B2 (en) * | 1992-02-19 | 2005-11-15 | 8×8, Inc. | Programmable architecture and methods for motion estimation |
US6215910B1 (en) * | 1996-03-28 | 2001-04-10 | Microsoft Corporation | Table-based compression with embedded coding |
US6571016B1 (en) * | 1997-05-05 | 2003-05-27 | Microsoft Corporation | Intra compression of pixel blocks using predicted mean |
WO1998015125A1 (en) * | 1996-10-01 | 1998-04-09 | Siemens Aktiengesellschaft | Method and arrangement for vector quantization and for reverse vector quantization of a digitized image |
US6332030B1 (en) * | 1998-01-15 | 2001-12-18 | The Regents Of The University Of California | Method for embedding and extracting digital data in images and video |
US6327392B1 (en) * | 1999-01-28 | 2001-12-04 | Sharp Laboratories Of America, Inc. | Method of visual progressive coding |
US7309829B1 (en) | 1998-05-15 | 2007-12-18 | Ludwig Lester F | Layered signal processing for individual and group output of multi-channel electronic musical instruments |
US6610917B2 (en) * | 1998-05-15 | 2003-08-26 | Lester F. Ludwig | Activity indication, external source, and processing loop provisions for driven vibrating-element environments |
US20050120870A1 (en) * | 1998-05-15 | 2005-06-09 | Ludwig Lester F. | Envelope-controlled dynamic layering of audio signal processing and synthesis for music applications |
US6470052B1 (en) * | 1998-06-29 | 2002-10-22 | Xerox Corporation | HVQ compression combined with orthogonal rotation |
DE69934939T2 (en) | 1998-06-29 | 2007-10-18 | Xerox Corp. | Compression of boundaries between images |
US6731807B1 (en) * | 1998-09-11 | 2004-05-04 | Intel Corporation | Method of compressing and/or decompressing a data set using significance mapping |
US6415326B1 (en) | 1998-09-15 | 2002-07-02 | Microsoft Corporation | Timeline correlation between multiple timeline-altered media streams |
US6622171B2 (en) * | 1998-09-15 | 2003-09-16 | Microsoft Corporation | Multimedia timeline modification in networked client/server systems |
US6643405B1 (en) * | 1998-09-30 | 2003-11-04 | Canon Kabushiki Kaisha | Image processing apparatus and method and computer-readable storage medium |
JP2000115558A (en) * | 1998-10-08 | 2000-04-21 | Mitsubishi Electric Corp | Color characteristic description device, color management device, image converter and color correction method |
JP2000197054A (en) * | 1998-12-24 | 2000-07-14 | Hudson Soft Co Ltd | Dynamic image encoding method, recording medium with program thereof recorded therein and apparatus |
US6693962B1 (en) * | 1999-02-01 | 2004-02-17 | Thomson Licensing S.A. | Process to extract regions of homogeneous texture in a digital picture |
US6778697B1 (en) | 1999-02-05 | 2004-08-17 | Samsung Electronics Co., Ltd. | Color image processing method and apparatus thereof |
DE60034554T2 (en) * | 1999-02-05 | 2007-12-27 | Samsung Electronics Co., Ltd., Suwon | METHOD AND DEVICE FOR MACHINING COLOR PICTURES |
GB9903074D0 (en) * | 1999-02-12 | 1999-03-31 | Silver Fox Ltd | Computer system for producing series of sequential images for labelling |
US6842724B1 (en) * | 1999-04-08 | 2005-01-11 | Lucent Technologies Inc. | Method and apparatus for reducing start-up delay in data packet-based network streaming applications |
JP2000305822A (en) * | 1999-04-26 | 2000-11-02 | Denso Corp | Device for database management and device for database extraction, and method for database management and method for database extraction |
US7313808B1 (en) | 1999-07-08 | 2007-12-25 | Microsoft Corporation | Browsing continuous multimedia content |
US7293280B1 (en) | 1999-07-08 | 2007-11-06 | Microsoft Corporation | Skimming continuous multimedia content |
US7170941B2 (en) * | 1999-08-13 | 2007-01-30 | Patapsco Designs Inc. | Temporal compression |
US6621935B1 (en) * | 1999-12-03 | 2003-09-16 | Microsoft Corporation | System and method for robust image representation over error-prone channels |
US7149359B1 (en) | 1999-12-16 | 2006-12-12 | Microsoft Corporation | Searching and recording media streams |
US6928655B1 (en) * | 1999-12-16 | 2005-08-09 | Microsoft Corporation | Live presentation searching |
CN101383909B (en) * | 2000-01-26 | 2013-03-20 | 索尼公司 | Information processing apparatus and processing method |
US6868440B1 (en) * | 2000-02-04 | 2005-03-15 | Microsoft Corporation | Multi-level skimming of multimedia content using playlists |
US7237254B1 (en) | 2000-03-29 | 2007-06-26 | Microsoft Corporation | Seamless switching between different playback speeds of time-scale modified data streams |
US6985966B1 (en) | 2000-03-29 | 2006-01-10 | Microsoft Corporation | Resynchronizing globally unsynchronized multimedia streams |
US7302490B1 (en) | 2000-05-03 | 2007-11-27 | Microsoft Corporation | Media file format to support switching between multiple timeline-altered media streams |
US7039241B1 (en) * | 2000-08-11 | 2006-05-02 | Ati Technologies, Inc. | Method and apparatus for compression and decompression of color data |
US6678422B1 (en) * | 2000-08-30 | 2004-01-13 | National Semiconductor Corporation | Method and apparatus for image data compression with low memory requirement |
US6683994B1 (en) * | 2000-11-28 | 2004-01-27 | Xerox Corporation | HVQ-based filtering method |
US7069545B2 (en) * | 2000-12-29 | 2006-06-27 | Intel Corporation | Quantization and compression for computation reuse |
KR100359118B1 (en) * | 2001-03-20 | 2002-11-04 | 주식회사 아름테크놀로지 | Lossless data compression method for uniform entropy data |
WO2002086821A1 (en) * | 2001-04-19 | 2002-10-31 | Kabushiki Kaisha Toshiba | Image processing method and image processing device |
US6691132B2 (en) * | 2001-05-16 | 2004-02-10 | Reengineering Llc | Semantic encoding and compression of database tables |
US20020191698A1 (en) * | 2001-06-15 | 2002-12-19 | Solidstreaming, Inc. | Video data CODEC system with low computational complexity |
US6807312B2 (en) * | 2001-07-13 | 2004-10-19 | Sharp Laboratories Of America, Inc. | Robust codebooks for vector quantization |
AU2002343740A1 (en) * | 2001-11-16 | 2003-06-10 | California Institute Of Technology | Data compression method and system |
CN101448162B (en) | 2001-12-17 | 2013-01-02 | 微软公司 | Method for processing video image |
US7200275B2 (en) | 2001-12-17 | 2007-04-03 | Microsoft Corporation | Skip macroblock coding |
US7043514B1 (en) * | 2002-03-01 | 2006-05-09 | Microsoft Corporation | System and method adapted to facilitate dimensional transform |
FI113931B (en) * | 2002-05-23 | 2004-06-30 | Oplayo Oy | Method and system for constructing the HCVQ vector library |
CA2388358A1 (en) * | 2002-05-31 | 2003-11-30 | Voiceage Corporation | A method and device for multi-rate lattice vector quantization |
US7016547B1 (en) | 2002-06-28 | 2006-03-21 | Microsoft Corporation | Adaptive entropy encoding/decoding for screen capture content |
US7433824B2 (en) * | 2002-09-04 | 2008-10-07 | Microsoft Corporation | Entropy coding by adapting coding between level and run-length/level modes |
EP2006840B1 (en) | 2002-09-04 | 2012-07-04 | Microsoft Corporation | Entropy coding by adapting coding between level and run-length/level modes |
US7554980B1 (en) * | 2002-10-18 | 2009-06-30 | Alcatel Lucent | Packet classification using relevance scoring |
US7558441B2 (en) * | 2002-10-24 | 2009-07-07 | Canon Kabushiki Kaisha | Resolution conversion upon hierarchical coding and decoding |
US7106901B2 (en) * | 2003-01-15 | 2006-09-12 | Xerox Corporation | Method and apparatus for image classification |
US8705613B2 (en) * | 2003-06-26 | 2014-04-22 | Sony Corporation | Adaptive joint source channel coding |
US20050013589A1 (en) * | 2003-07-14 | 2005-01-20 | Microsoft Corporation | Adding recording functionality to a media player |
US10554985B2 (en) | 2003-07-18 | 2020-02-04 | Microsoft Technology Licensing, Llc | DC coefficient signaling at small quantization step sizes |
US7830963B2 (en) * | 2003-07-18 | 2010-11-09 | Microsoft Corporation | Decoding jointly coded transform type and subblock pattern information |
US7724827B2 (en) * | 2003-09-07 | 2010-05-25 | Microsoft Corporation | Multi-layer run level encoding and decoding |
US7688894B2 (en) * | 2003-09-07 | 2010-03-30 | Microsoft Corporation | Scan patterns for interlaced video content |
US7782954B2 (en) * | 2003-09-07 | 2010-08-24 | Microsoft Corporation | Scan patterns for progressive video content |
US8582640B2 (en) * | 2003-12-16 | 2013-11-12 | Sony Corporation | Adaptive joint source channel coding |
KR100868716B1 (en) | 2004-04-21 | 2008-11-13 | 슬립스트림 데이터 인크. | Method, system and software product for color image encoding |
US7525552B2 (en) * | 2004-04-21 | 2009-04-28 | Slipstream Data Inc. | Method, system and software product for color image encoding |
US7697759B2 (en) * | 2004-05-11 | 2010-04-13 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Split-remerge method for eliminating processing window artifacts in recursive hierarchical segmentation |
JP2007537815A (en) * | 2004-05-18 | 2007-12-27 | コーニンクレッカ フィリップス エレクトロニクス エヌ ヴィ | An image processing system for automatic segmentation of 3D tree-like tubular surfaces of objects using 3D deformed mesh models |
KR100654436B1 (en) * | 2004-07-07 | 2006-12-06 | 삼성전자주식회사 | Method for video encoding and decoding, and video encoder and decoder |
KR100657919B1 (en) * | 2004-12-13 | 2006-12-14 | 삼성전자주식회사 | Apparatus and method for intra prediction of an image data, apparatus and method for encoding of an image data, apparatus and method for intra prediction compensation of an image data, apparatus and method for decoding of an image data |
US7953155B2 (en) * | 2004-12-16 | 2011-05-31 | Tut Systems, Inc. | Methods of selecting an encoding mode |
US20060262860A1 (en) * | 2005-02-23 | 2006-11-23 | Chou Jim C | Macroblock adaptive frame/field coding architecture for scalable coding |
US20060204086A1 (en) * | 2005-03-10 | 2006-09-14 | Ullas Gargi | Compression of palettized images |
US7434146B1 (en) * | 2005-05-06 | 2008-10-07 | Helwett-Packard Development Company, L.P. | Denoising and error correction for finite input, general output channel |
KR100716999B1 (en) * | 2005-06-03 | 2007-05-10 | 삼성전자주식회사 | Method for intra prediction using the symmetry of video, method and apparatus for encoding and decoding video using the same |
NZ564059A (en) * | 2005-06-03 | 2010-10-29 | Commw Of Australia | Messaging method |
US7599840B2 (en) * | 2005-07-15 | 2009-10-06 | Microsoft Corporation | Selectively using multiple entropy models in adaptive coding and decoding |
US7684981B2 (en) * | 2005-07-15 | 2010-03-23 | Microsoft Corporation | Prediction of spectral coefficients in waveform coding and decoding |
US7693709B2 (en) | 2005-07-15 | 2010-04-06 | Microsoft Corporation | Reordering coefficients for waveform coding or decoding |
US7565018B2 (en) * | 2005-08-12 | 2009-07-21 | Microsoft Corporation | Adaptive coding and decoding of wide-range coefficients |
US7933337B2 (en) * | 2005-08-12 | 2011-04-26 | Microsoft Corporation | Prediction of transform coefficients for image compression |
US8599925B2 (en) * | 2005-08-12 | 2013-12-03 | Microsoft Corporation | Efficient coding and decoding of transform blocks |
US7464316B2 (en) * | 2005-08-26 | 2008-12-09 | Agere Systems Inc. | Modified branch metric calculator to reduce interleaver memory and improve performance in a fixed-point turbo decoder |
US7587314B2 (en) * | 2005-08-29 | 2009-09-08 | Nokia Corporation | Single-codebook vector quantization for multiple-rate applications |
US8325748B2 (en) * | 2005-09-16 | 2012-12-04 | Oracle International Corporation | Fast vector quantization with topology learning |
CN100472556C (en) * | 2005-10-09 | 2009-03-25 | 欧姆龙株式会社 | Testing apparatus and method for special object |
FR2893432B1 (en) * | 2005-11-16 | 2008-01-04 | Atmel Corp | VECTOR QUANTIFIER BASED ON SPACE DICHOTOMY WITH N DIMENSIONS |
US7778472B2 (en) * | 2006-03-27 | 2010-08-17 | Qualcomm Incorporated | Methods and systems for significance coefficient coding in video compression |
JP4679425B2 (en) | 2006-04-20 | 2011-04-27 | 株式会社東芝 | Image processing apparatus, image processing method, and program |
US20070286277A1 (en) * | 2006-06-13 | 2007-12-13 | Chen Xuemin Sherman | Method and system for video compression using an iterative encoding algorithm |
US8176016B1 (en) * | 2006-11-17 | 2012-05-08 | At&T Intellectual Property Ii, L.P. | Method and apparatus for rapid identification of column heterogeneity |
US8184710B2 (en) * | 2007-02-21 | 2012-05-22 | Microsoft Corporation | Adaptive truncation of transform coefficient data in a transform-based digital media codec |
KR101403338B1 (en) * | 2007-03-23 | 2014-06-09 | 삼성전자주식회사 | Method and apparatus for image encoding, decoding |
US7957600B2 (en) * | 2007-05-08 | 2011-06-07 | Arris Group, Inc. | Methods and systems for rate-distortion optimized quantization of transform blocks in block transform video coding |
US7774205B2 (en) * | 2007-06-15 | 2010-08-10 | Microsoft Corporation | Coding of sparse digital media spectral data |
US8295624B2 (en) | 2007-12-03 | 2012-10-23 | Ecole De Technologie Superieure | Method and system for generating a quality prediction table for quality-aware transcoding of digital images |
WO2009055896A1 (en) | 2007-11-02 | 2009-05-07 | Vantrix Corporation | System and method for predicting the file size of images subject to transformation by scaling and a change of quality-controlling parameters |
US9019237B2 (en) * | 2008-04-06 | 2015-04-28 | Lester F. Ludwig | Multitouch parameter and gesture user interface employing an LED-array tactile sensor that can also operate as a display |
US8179974B2 (en) | 2008-05-02 | 2012-05-15 | Microsoft Corporation | Multi-level representation of reordered transform coefficients |
US8169414B2 (en) | 2008-07-12 | 2012-05-01 | Lim Seung E | Control of electronic games via finger angle using a high dimensional touchpad (HDTP) touch user interface |
US8345014B2 (en) | 2008-07-12 | 2013-01-01 | Lester F. Ludwig | Control of the operating system on a computing device via finger angle using a high dimensional touchpad (HDTP) touch user interface |
US8604364B2 (en) * | 2008-08-15 | 2013-12-10 | Lester F. Ludwig | Sensors, algorithms and applications for a high dimensional touchpad |
US8406307B2 (en) | 2008-08-22 | 2013-03-26 | Microsoft Corporation | Entropy coding/decoding of hierarchically organized data |
US8175265B2 (en) * | 2008-09-02 | 2012-05-08 | Apple Inc. | Systems and methods for implementing block cipher algorithms on attacker-controlled systems |
US8503527B2 (en) | 2008-10-03 | 2013-08-06 | Qualcomm Incorporated | Video coding with large macroblocks |
TWI384767B (en) * | 2008-11-21 | 2013-02-01 | Univ Nat Chiao Tung | Method, apparatus and computer program product for partition a codebook and selection a precoding codeword from the codebook |
US8300961B2 (en) * | 2008-12-12 | 2012-10-30 | Ecole De Technologie Superieure | Method and system for low complexity transcoding of images with near optimal quality |
US8170346B2 (en) | 2009-03-14 | 2012-05-01 | Ludwig Lester F | High-performance closed-form single-scan calculation of oblong-shape rotation angles from binary images of arbitrary size using running sums |
US8351723B2 (en) * | 2009-03-25 | 2013-01-08 | Accusoft-Pegasus | Methods and apparatus for generating and using visual error weights |
CN101577551A (en) * | 2009-05-27 | 2009-11-11 | 华为技术有限公司 | Method and device for generating lattice vector quantization codebook |
AU2015202118B2 (en) * | 2009-08-17 | 2015-06-11 | Samsung Electronics Co., Ltd. | Method and apparatus for encoding video, and method and apparatus for decoding video |
KR101510108B1 (en) * | 2009-08-17 | 2015-04-10 | 삼성전자주식회사 | Method and apparatus for encoding video, and method and apparatus for decoding video |
US20110066933A1 (en) | 2009-09-02 | 2011-03-17 | Ludwig Lester F | Value-driven visualization primitives for spreadsheets, tabular data, and advanced spreadsheet visualization |
US20110055722A1 (en) * | 2009-09-02 | 2011-03-03 | Ludwig Lester F | Data Visualization Environment with DataFlow Processing, Web, Collaboration, Advanced User Interfaces, and Spreadsheet Visualization |
EP2488107B1 (en) | 2009-10-12 | 2017-03-08 | Acist Medical Systems, Inc. | Intravascular ultrasound system for co-registered imaging |
US20110202934A1 (en) * | 2010-02-12 | 2011-08-18 | Ludwig Lester F | Window manger input focus control for high dimensional touchpad (htpd), advanced mice, and other multidimensional user interfaces |
US10146427B2 (en) | 2010-03-01 | 2018-12-04 | Nri R&D Patent Licensing, Llc | Curve-fitting approach to high definition touch pad (HDTP) parameter extraction |
US9626023B2 (en) | 2010-07-09 | 2017-04-18 | Lester F. Ludwig | LED/OLED array approach to integrated display, lensless-camera, and touch-screen user interface devices and associated processors |
US9632344B2 (en) | 2010-07-09 | 2017-04-25 | Lester F. Ludwig | Use of LED or OLED array to implement integrated combinations of touch screen tactile, touch gesture sensor, color image display, hand-image gesture sensor, document scanner, secure optical data exchange, and fingerprint processing capabilities |
US8754862B2 (en) | 2010-07-11 | 2014-06-17 | Lester F. Ludwig | Sequential classification recognition of gesture primitives and window-based parameter smoothing for high dimensional touchpad (HDTP) user interfaces |
US9950256B2 (en) | 2010-08-05 | 2018-04-24 | Nri R&D Patent Licensing, Llc | High-dimensional touchpad game controller with multiple usage and networking modalities |
US8712930B1 (en) | 2010-08-09 | 2014-04-29 | Google Inc. | Encoding digital content based on models for predicting similarity between exemplars |
US20120204577A1 (en) | 2011-02-16 | 2012-08-16 | Ludwig Lester F | Flexible modular hierarchical adaptively controlled electronic-system cooling and energy harvesting for IC chip packaging, printed circuit boards, subsystems, cages, racks, IT rooms, and data centers using quantum and classical thermoelectric materials |
US9442652B2 (en) | 2011-03-07 | 2016-09-13 | Lester F. Ludwig | General user interface gesture lexicon and grammar frameworks for multi-touch, high dimensional touch pad (HDTP), free-space camera, and other user interfaces |
US8515193B1 (en) * | 2011-04-08 | 2013-08-20 | Google Inc. | Image compression using exemplar dictionary based on hierarchical clustering |
US8542934B2 (en) | 2011-05-18 | 2013-09-24 | Microsoft Corporation | Transform space difference compression |
US9052772B2 (en) | 2011-08-10 | 2015-06-09 | Lester F. Ludwig | Heuristics for 3D and 6D touch gesture touch parameter calculations for high-dimensional touch parameter (HDTP) user interfaces |
US10430066B2 (en) | 2011-12-06 | 2019-10-01 | Nri R&D Patent Licensing, Llc | Gesteme (gesture primitive) recognition for advanced touch user interfaces |
US9823781B2 (en) | 2011-12-06 | 2017-11-21 | Nri R&D Patent Licensing, Llc | Heterogeneous tactile sensing via multiple sensor types |
US10609405B2 (en) | 2013-03-18 | 2020-03-31 | Ecole De Technologie Superieure | Optimal signal encoding based on experimental data |
US9661331B2 (en) | 2013-03-18 | 2017-05-23 | Vantrix Corporation | Method and apparatus for signal encoding realizing optimal fidelity |
US9338450B2 (en) | 2013-03-18 | 2016-05-10 | Ecole De Technologie Superieure | Method and apparatus for signal encoding producing encoded signals of high fidelity at minimal sizes |
US9141885B2 (en) * | 2013-07-29 | 2015-09-22 | Adobe Systems Incorporated | Visual pattern recognition in an image |
EP3055709B1 (en) | 2013-10-07 | 2018-12-26 | Acist Medical Systems, Inc. | Signal processing for intravascular imaging |
DE102013021599B3 (en) * | 2013-12-19 | 2014-12-31 | Wolfgang Klippel | Arrangement and method for reducing the nonlinear distortion of an AD converter |
US11350015B2 (en) | 2014-01-06 | 2022-05-31 | Panamorph, Inc. | Image processing system and method |
WO2015103646A1 (en) | 2014-01-06 | 2015-07-09 | Panamorph, Inc. | Image processing system and method |
US10909661B2 (en) | 2015-10-08 | 2021-02-02 | Acist Medical Systems, Inc. | Systems and methods to reduce near-field artifacts |
US10653393B2 (en) | 2015-10-08 | 2020-05-19 | Acist Medical Systems, Inc. | Intravascular ultrasound imaging with frequency selective imaging methods and systems |
US11369337B2 (en) | 2015-12-11 | 2022-06-28 | Acist Medical Systems, Inc. | Detection of disturbed blood flow |
US10275881B2 (en) | 2015-12-31 | 2019-04-30 | Val-Chum, Limited Partnership | Semi-automated image segmentation system and method |
CN105611303B (en) * | 2016-03-07 | 2019-04-09 | 京东方科技集团股份有限公司 | Image compression system, decompression systems, training method and device, display device |
CN109478332B (en) | 2016-05-16 | 2023-04-25 | 阿西斯特医药系统公司 | Motion-based image segmentation system and method |
EP3386198A1 (en) * | 2017-04-07 | 2018-10-10 | Thomson Licensing | Method and device for predictive picture encoding and decoding |
US11423312B2 (en) | 2018-05-14 | 2022-08-23 | Samsung Electronics Co., Ltd | Method and apparatus for universal pruning and compression of deep convolutional neural networks under joint sparsity constraints |
KR20210134391A (en) * | 2019-03-12 | 2021-11-09 | 후아웨이 테크놀러지 컴퍼니 리미티드 | Coding and decoding patch data units for point cloud coding |
US11024034B2 (en) | 2019-07-02 | 2021-06-01 | Acist Medical Systems, Inc. | Image segmentation confidence determination |
US10904049B1 (en) | 2019-07-11 | 2021-01-26 | Stmicroelectronics (Research & Development) Limited | Time domain discrete transform computation |
US11681708B2 (en) | 2019-12-26 | 2023-06-20 | Snowflake Inc. | Indexed regular expression search with N-grams |
US11372860B2 (en) | 2019-12-26 | 2022-06-28 | Snowflake Inc. | Processing techniques for queries where predicate values are unknown until runtime |
US11016975B1 (en) | 2019-12-26 | 2021-05-25 | Snowflake Inc. | Scan set pruning for pattern matching queries |
US10769150B1 (en) | 2019-12-26 | 2020-09-08 | Snowflake Inc. | Pruning indexes to enhance database query processing |
US10997179B1 (en) | 2019-12-26 | 2021-05-04 | Snowflake Inc. | Pruning index for optimization of pattern matching queries |
US11567939B2 (en) | 2019-12-26 | 2023-01-31 | Snowflake Inc. | Lazy reassembling of semi-structured data |
US11308090B2 (en) | 2019-12-26 | 2022-04-19 | Snowflake Inc. | Pruning index to support semi-structured data types |
CN111726629B (en) * | 2020-06-09 | 2022-02-11 | 绍兴图信科技有限公司 | SMVQ compressed data hiding method based on multiple linear regression |
TWI780696B (en) * | 2021-05-10 | 2022-10-11 | 創鑫智慧股份有限公司 | Look-up table compression method and look-up table reading method and computation equipment, host and device thereof |
CN115391298A (en) * | 2021-05-25 | 2022-11-25 | 戴尔产品有限公司 | Content-based dynamic hybrid data compression |
US20230289458A1 (en) * | 2022-03-11 | 2023-09-14 | Intuit Inc. | Seed generation for electronic data perturbation |
US11880369B1 (en) | 2022-11-21 | 2024-01-23 | Snowflake Inc. | Pruning data based on state of top K operator |
Family Cites Families (57)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE3704777C1 (en) | 1987-02-16 | 1988-04-07 | Ant Nachrichtentech | Method of transmitting and playing back television picture sequences |
EP0387051B1 (en) | 1989-03-10 | 1997-08-06 | Canon Kabushiki Kaisha | Method and apparatus for coding image information |
US5060285A (en) | 1989-05-19 | 1991-10-22 | Gte Laboratories Incorporated | Hierarchical variable block size address-vector quantization using inter-block correlation |
US4987480A (en) | 1989-07-11 | 1991-01-22 | Massachusetts Institute Of Technology | Multiscale coding of images |
FR2651399B1 (en) | 1989-08-29 | 1996-05-15 | Thomson Consumer Electronics | METHOD AND DEVICE FOR ESTIMATING AND HIERARCHIZED CODING OF THE MOTION OF IMAGE SEQUENCES. |
US5144688A (en) | 1990-03-23 | 1992-09-01 | Board Of Regents, The University Of Texas System | Method and apparatus for visual pattern image coding |
US5155594A (en) | 1990-05-11 | 1992-10-13 | Picturetel Corporation | Hierarchical encoding method and apparatus employing background references for efficiently communicating image sequences |
US5068724A (en) | 1990-06-15 | 1991-11-26 | General Instrument Corporation | Adaptive motion compensation for digital television |
US5260783A (en) * | 1991-02-21 | 1993-11-09 | Gte Laboratories Incorporated | Layered DCT video coder for packet switched ATM networks |
SE469866B (en) | 1991-04-12 | 1993-09-27 | Dv Sweden Ab | Method for estimating motion content in video signals |
US5144425A (en) | 1991-08-26 | 1992-09-01 | General Electric Company | Apparatus for hierarchically dividing video signals |
EP0533195A2 (en) | 1991-09-20 | 1993-03-24 | Sony Corporation | Picture signal encoding and/or decoding apparatus |
US5414469A (en) | 1991-10-31 | 1995-05-09 | International Business Machines Corporation | Motion video compression system with multiresolution features |
US5227878A (en) | 1991-11-15 | 1993-07-13 | At&T Bell Laboratories | Adaptive coding and decoding of frames and fields of video |
US5371544A (en) | 1992-02-07 | 1994-12-06 | At&T Corp. | Geometric vector quantization |
US6215425B1 (en) | 1992-02-19 | 2001-04-10 | Netergy Networks, Inc. | Deblocking filter for encoder/decoder arrangement and method with divergence reduction |
US5325126A (en) | 1992-04-01 | 1994-06-28 | Intel Corporation | Method and apparatus for real time compression and decompression of a digital motion video signal |
KR0166716B1 (en) | 1992-06-18 | 1999-03-20 | 강진구 | Encoding and decoding method and apparatus by using block dpcm |
US5596659A (en) | 1992-09-01 | 1997-01-21 | Apple Computer, Inc. | Preprocessing and postprocessing for vector quantization |
US5467134A (en) | 1992-12-22 | 1995-11-14 | Microsoft Corporation | Method and system for compressing video data |
US5544286A (en) | 1993-01-29 | 1996-08-06 | Microsoft Corporation | Digital video data compression technique |
WO1994018799A1 (en) | 1993-02-03 | 1994-08-18 | Qualcomm Incorporated | Interframe video encoding and decoding system |
US5592228A (en) * | 1993-03-04 | 1997-01-07 | Kabushiki Kaisha Toshiba | Video encoder using global motion estimation and polygonal patch motion estimation |
KR960015395B1 (en) | 1993-04-09 | 1996-11-11 | 대우전자 주식회사 | Motion vector detecting apparatus |
US5442400A (en) | 1993-04-29 | 1995-08-15 | Rca Thomson Licensing Corporation | Error concealment apparatus for MPEG-like video data |
US5585852A (en) | 1993-06-16 | 1996-12-17 | Intel Corporation | Processing video signals for scalable video playback using independently encoded component-plane bands |
KR0128860B1 (en) | 1993-07-16 | 1998-04-10 | 배순훈 | Encoding apparatus in phone system with lowbit rate |
JP3590996B2 (en) | 1993-09-30 | 2004-11-17 | ソニー株式会社 | Hierarchical encoding and decoding apparatus for digital image signal |
US5473379A (en) | 1993-11-04 | 1995-12-05 | At&T Corp. | Method and apparatus for improving motion compensation in digital video coding |
EP0665688A3 (en) | 1993-12-29 | 1995-11-02 | Toshiba Kk | Video data arranging method and video data encoding/decoding apparatus. |
US5521988A (en) | 1994-04-05 | 1996-05-28 | Gte Laboratories Incorporated | Vector transform coder with multi-layered codebooks and dynamic bit allocation |
JP3013698B2 (en) * | 1994-04-20 | 2000-02-28 | 松下電器産業株式会社 | Vector quantization encoding device and decoding device |
US5537155A (en) | 1994-04-29 | 1996-07-16 | Motorola, Inc. | Method for estimating motion in a video sequence |
US5555244A (en) | 1994-05-19 | 1996-09-10 | Integrated Network Corporation | Scalable multimedia network |
US5604867A (en) | 1994-07-22 | 1997-02-18 | Network Peripherals | System for transmitting data between bus and network having device comprising first counter for providing transmitting rate and second counter for limiting frames exceeding rate |
US5655140A (en) | 1994-07-22 | 1997-08-05 | Network Peripherals | Apparatus for translating frames of data transferred between heterogeneous local area networks |
US5602589A (en) | 1994-08-19 | 1997-02-11 | Xerox Corporation | Video image compression using weighted wavelet hierarchical vector quantization |
US5802213A (en) | 1994-10-18 | 1998-09-01 | Intel Corporation | Encoding video signals using local quantization levels |
US5623312A (en) | 1994-12-22 | 1997-04-22 | Lucent Technologies Inc. | Compressed-domain bit rate reduction system |
US5623313A (en) | 1995-09-22 | 1997-04-22 | Tektronix, Inc. | Fractional pixel motion estimation of video signals |
US5959673A (en) | 1995-10-05 | 1999-09-28 | Microsoft Corporation | Transform coding of dense motion vector fields for frame and object based video coding applications |
JP2798035B2 (en) | 1996-01-17 | 1998-09-17 | 日本電気株式会社 | Motion compensated inter-frame prediction method using adaptive motion vector interpolation |
US5799113A (en) | 1996-01-19 | 1998-08-25 | Microsoft Corporation | Method for expanding contracted video images |
US5764814A (en) | 1996-03-22 | 1998-06-09 | Microsoft Corporation | Representation and encoding of general arbitrary shapes |
US5778098A (en) | 1996-03-22 | 1998-07-07 | Microsoft Corporation | Sprite coding |
US6571016B1 (en) | 1997-05-05 | 2003-05-27 | Microsoft Corporation | Intra compression of pixel blocks using predicted mean |
US6215910B1 (en) | 1996-03-28 | 2001-04-10 | Microsoft Corporation | Table-based compression with embedded coding |
KR100403077B1 (en) | 1996-05-28 | 2003-10-30 | 마쯔시다덴기산교 가부시키가이샤 | Image predictive decoding apparatus and method thereof, and image predictive cording apparatus and method thereof |
US5952943A (en) | 1996-10-11 | 1999-09-14 | Intel Corporation | Encoding image data for decode rate control |
US6281942B1 (en) | 1997-08-11 | 2001-08-28 | Microsoft Corporation | Spatial and temporal filtering mechanism for digital motion video signals |
US6339656B1 (en) | 1997-12-25 | 2002-01-15 | Matsushita Electric Industrial Co., Ltd. | Moving picture encoding decoding processing apparatus |
US5946043A (en) | 1997-12-31 | 1999-08-31 | Microsoft Corporation | Video coding using adaptive coding of block parameters for coded/uncoded blocks |
US6236764B1 (en) | 1998-11-30 | 2001-05-22 | Equator Technologies, Inc. | Image processing circuit and method for reducing a difference between pixel values across an image boundary |
US6563953B2 (en) | 1998-11-30 | 2003-05-13 | Microsoft Corporation | Predictive image compression using a single variable length code for both the luminance and chrominance blocks for each macroblock |
US6300888B1 (en) | 1998-12-14 | 2001-10-09 | Microsoft Corporation | Entrophy code mode switching for frequency-domain audio coding |
US6654419B1 (en) | 2000-04-28 | 2003-11-25 | Sun Microsystems, Inc. | Block-based, adaptive, lossless video coder |
KR20050061762A (en) | 2003-12-18 | 2005-06-23 | 학교법인 대양학원 | Method of encoding mode determination and motion estimation, and encoding apparatus |
-
1996
- 1996-03-28 US US08/623,299 patent/US6215910B1/en not_active Expired - Lifetime
-
1997
- 1997-03-26 WO PCT/US1997/004879 patent/WO1997036376A1/en not_active Application Discontinuation
- 1997-03-26 JP JP9534591A patent/JP2000507754A/en active Pending
- 1997-03-26 EP EP97917006A patent/EP0890222A1/en not_active Withdrawn
- 1997-03-26 AU AU25472/97A patent/AU2547297A/en not_active Abandoned
- 1997-06-30 US US08/884,790 patent/US6349152B1/en not_active Expired - Lifetime
- 1997-06-30 US US08/884,792 patent/US6205256B1/en not_active Expired - Lifetime
- 1997-06-30 US US08/884,791 patent/US6360019B1/en not_active Expired - Lifetime
- 1997-06-30 US US08/885,159 patent/US6154572A/en not_active Expired - Lifetime
-
2003
- 2003-03-31 US US10/403,780 patent/US7162091B2/en not_active Expired - Lifetime
Non-Patent Citations (6)
Title |
---|
CHADDHA ; GUPTA: "a frame-work for live multicast of video streams over the internet", PROCEEDINGS OF 3RD IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING, vol. 1, 16 September 1996 (1996-09-16) - 19 September 1996 (1996-09-19), NEW YORK ,USA, pages 1 - 4, XP002036749 * |
CHADDHA ; WALL: "an end-to-end software only scalable delivery system", PROCEEDINGS NETWORKS AND OPERATING SYSTEM SUPPORT FOR DIGITAL AUDIO AND VIDEO, 19 April 1995 (1995-04-19) - 21 April 1995 (1995-04-21), DURHAM,N.H.,USA, pages 130 - 141, XP002036750 * |
CHADDHA ET AL: "hierarchical quantization of perceptually weighted block transforms", PROCEEDINGS DATA COMPRESSION CONFERENCE, 28 March 1995 (1995-03-28) - 30 March 1995 (1995-03-30), LOS ALAMITOS,CA.,USA, pages 3 - 12, XP002036751 * |
GERSHO ET AL: "vector quantization and signal compression", 1991, KLUWER, BOSTON,US, XP002036752 * |
KATSAVOUNIDIS; KUO: "hierarchical weighted vector quantization (hwvq) for embedded wavelet coding", PROCEEDINGS OF THE SPIE-THE INTERNATIONAL SOCIETY FOR OPTICAL ENGINEERING,1996, vol. 2727, 17 March 1996 (1996-03-17) - 20 March 1996 (1996-03-20), ORLANDO,FL,US, pages 548 - 549, XP002036748 * |
YANG ET AL: "motion-compensated wavelet transform coder for very low bit-rate visual telephony", SIGNAL PROCESSING : IMAGE COMMUNICATION, vol. 7, 1995, AMSTERDAM, pages 581 - 592, XP002036754 * |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6130911A (en) * | 1997-11-21 | 2000-10-10 | Sharp Laboratories Of America, Inc. | Method and apparatus for compressing reference frames in an interframe video codec |
WO1999027715A1 (en) * | 1997-11-21 | 1999-06-03 | Sharp Laboratories Of America, Inc. | Method and apparatus for compressing reference frames in an interframe video codec |
US6272180B1 (en) | 1997-11-21 | 2001-08-07 | Sharp Laboratories Of America, Inc. | Compression and decompression of reference frames in a video decoder |
US6345126B1 (en) | 1998-01-29 | 2002-02-05 | Xerox Corporation | Method for transmitting data using an embedded bit stream produced in a hierarchical table-lookup vector quantizer |
EP0933946A2 (en) * | 1998-01-29 | 1999-08-04 | Xerox Corporation | A method for transmitting data using an embedded bit stream produced in a hierarchical tablelookup vector quantizer |
US6671416B2 (en) | 1998-01-29 | 2003-12-30 | Xerox Corporation | Method for transmitting data using an embedded bit stream produced in a hierarchical table-lookup vector quantizer |
EP0933946A3 (en) * | 1998-01-29 | 1999-12-08 | Xerox Corporation | A method for transmitting data using an embedded bit stream produced in a hierarchical tablelookup vector quantizer |
JP2010181889A (en) * | 1998-08-24 | 2010-08-19 | Mindspeed Technologies Inc | Selection of scalar quantization (sq) and vector quantization (vq) for speech encoding |
EP1009167A3 (en) * | 1998-12-07 | 2002-01-02 | Xerox Corporation | Method and apparatus for electronic data compression |
EP1009167A2 (en) * | 1998-12-07 | 2000-06-14 | Xerox Corporation | Method and apparatus for electronic data compression |
WO2001031497A1 (en) * | 1999-10-22 | 2001-05-03 | Activesky, Inc. | An object oriented video system |
WO2002075929A1 (en) * | 2001-03-16 | 2002-09-26 | Coression Oy | A method for compressing data |
WO2006024077A2 (en) * | 2004-08-31 | 2006-03-09 | Media World Communications Limited | System and method for encoding and decoding video |
WO2006024077A3 (en) * | 2004-08-31 | 2008-02-21 | Media World Comm Ltd | System and method for encoding and decoding video |
US8219551B2 (en) | 2007-10-31 | 2012-07-10 | General Instrument Corporation | Decoding a hierarchical multi-layer data package |
US8788491B2 (en) | 2007-10-31 | 2014-07-22 | General Instrument Corporation | Decoding a hierarchical multi-layer data package |
CN102740325A (en) * | 2011-04-02 | 2012-10-17 | 华为技术有限公司 | Method, device for acquiring channel information and method, device for optimizing beam forming |
CN102740325B (en) * | 2011-04-02 | 2015-01-21 | 华为技术有限公司 | Method, device for acquiring channel information and method, device for optimizing beam forming |
CN102496368A (en) * | 2011-12-20 | 2012-06-13 | 重庆金美通信有限责任公司 | Improved vector quantization method |
Also Published As
Publication number | Publication date |
---|---|
US6205256B1 (en) | 2001-03-20 |
AU2547297A (en) | 1997-10-17 |
US6349152B1 (en) | 2002-02-19 |
US6360019B1 (en) | 2002-03-19 |
US7162091B2 (en) | 2007-01-09 |
EP0890222A1 (en) | 1999-01-13 |
US6154572A (en) | 2000-11-28 |
JP2000507754A (en) | 2000-06-20 |
US6215910B1 (en) | 2001-04-10 |
US20030185452A1 (en) | 2003-10-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US6154572A (en) | Table based compression with embedded coding | |
US20010017941A1 (en) | Method and apparatus for table-based compression with embedded coding | |
JP4554844B2 (en) | Method for generating a compressed digital image organized into a hierarchy in response to increasing visual quality levels and a method for controlling the rate of such compressed digital image | |
US5455874A (en) | Continuous-tone image compression | |
US8340450B2 (en) | Successively refinable lattice vector quantization | |
US6456744B1 (en) | Method and apparatus for video compression using sequential frame cellular automata transforms | |
JP3868014B2 (en) | Video signal encoding method and apparatus | |
US6885320B2 (en) | Apparatus and method for selecting length of variable length coding bit stream using neural network | |
Li et al. | Image compression using transformed vector quantization | |
US20030081852A1 (en) | Encoding method and arrangement | |
Markas et al. | Quad tree structures for image compression applications | |
Mohanta et al. | Image compression using different vector quantization algorithms and its comparison | |
US6330283B1 (en) | Method and apparatus for video compression using multi-state dynamical predictive systems | |
Kossentini et al. | Image coding using entropy-constrained residual vector quantization | |
Chaddha et al. | Constrained and recursive hierarchical table-lookup vector quantization | |
Kossentini et al. | Conditional entropy-constrained residual VQ with application to image coding | |
T Hashim et al. | Color image compression using DPCM with DCT, DWT and quadtree coding scheme | |
Bayazit et al. | Variable-length constrained-storage tree-structured vector quantization | |
Kossentini et al. | Finite-state residual vector quantization | |
JP2000041249A (en) | Visual progressive coding method | |
Barrilleaux et al. | Efficient vector quantization for color image encoding | |
Kumar et al. | An optimized block estimation based image compression and decompression algorithm | |
Somasundaram et al. | A pattern-based residual vector quantization (PBRVQ) algorithm for compressing images | |
Gifford et al. | Image coding using adaptive recursive interpolative DPCM | |
JP2914546B2 (en) | Singular value expansion image coding device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): AL AM AT AU AZ BA BB BG BR BY CA CH CN CU CZ DE DK EE ES FI GB GE GH HU IL IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MD MG MK MN MW MX NO NZ PL PT RO RU SD SE SG SI SK TJ TM TR TT UA UG US UZ VN YU AM AZ BY KG KZ MD RU TJ TM |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): GH KE LS MW SD SZ UG AT BE CH DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF |
|
DFPE | Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101) | ||
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
WWE | Wipo information: entry into national phase |
Ref document number: 1997917006 Country of ref document: EP |
|
WWP | Wipo information: published in national office |
Ref document number: 1997917006 Country of ref document: EP |
|
REG | Reference to national code |
Ref country code: DE Ref legal event code: 8642 |
|
WWW | Wipo information: withdrawn in national office |
Ref document number: 1997917006 Country of ref document: EP |
|
NENP | Non-entry into the national phase |
Ref country code: CA |