CN104834757A - Image semantic retrieval method and system - Google Patents
Image semantic retrieval method and system Download PDFInfo
- Publication number
- CN104834757A CN104834757A CN201510306112.4A CN201510306112A CN104834757A CN 104834757 A CN104834757 A CN 104834757A CN 201510306112 A CN201510306112 A CN 201510306112A CN 104834757 A CN104834757 A CN 104834757A
- Authority
- CN
- China
- Prior art keywords
- image
- semantic
- detection model
- retrieved
- described image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
Abstract
The invention provides an image semantic retrieval method and system. The image semantic retrieval method comprises the steps of extracting low-level visual features of an image, combining low-level visual features of different categories to form a comprehensive feature, building an image semantic detection model according to the comprehensive feature, perfecting the image semantic detection model through an incremental learning method, providing an image to be retrieved, and carrying out semantic annotation on the image to be retrieved through the perfected image semantic detection model. The image semantic retrieval method and system retrieve all image semantics in the image to be detected through the perfected image semantic detection model, and give a reasonable semantic annotation, wherein the image semantics comprise image semantics not learned by the image semantic detection model. Meanwhile, a display terminal can select different display schemes according to the semantic categories of images, and achieves intelligent display.
Description
Technical field
The present invention relates to technical field, particularly a kind of image semantic retrieving method and system thereof.
Background technology
Along with computing machine, the developing rapidly of multimedia and internet, internet there is the Digital image information resource of magnanimity.Picture search be in recent years growth rate the fastest classified search application, the picture search number of times of global a few large search engine is all doubled and redoubled.At present, image retrieval technologies has become the focus of research both at home and abroad.
Image retrieval technologies is mainly divided into text based image retrieval technologies (Text-based ImageRetrieval is called for short TBIR) and CBIR (Content-based Image Retrieval is called for short CBIR).Wherein, CBIR is the main flow of present image retrieval technique.
CBIR process comprises: first, automatically extracts bottom visual signature, as bottom visual signatures such as color, texture, profile and shapes from image; Then, according to described bottom Visual Feature Retrieval Process high-level semantics feature and design of graphics as Semantic detection model; Then, described image, semantic detection model is utilized to retrieve satisfactory result images.
In image, semantic detection model building process, due to manually enough semantic markers cannot be obtained, be therefore necessary that making full use of a large amount of unlabelled view data carrys out sophisticated image Semantic detection model.For this reason, usually never useful information is excavated to improve the Semantic detection accuracy rate of image, semantic detection model in marker samples.
At present, although the semi-supervised adaptive probability hypergraph model obtained based on incremental learning method can realize the multiple semantic retrieval of image, the foreign peoples's image, semantic do not learnt when model is set up cannot but be retrieved.Therefore, how to solve the problem that existing image semantic retrieving method cannot retrieve the foreign peoples's image, semantic do not learnt when image, semantic detection model is set up and become the technical matters that those skilled in the art need solution badly.
Summary of the invention
The object of the present invention is to provide a kind of image semantic retrieving method and system thereof, to solve the problem that existing image semantic retrieving method cannot retrieve the foreign peoples's image, semantic do not learnt when image, semantic detection model is set up.
For solving the problems of the technologies described above, the invention provides a kind of image semantic retrieving method, described image semantic retrieving method comprises:
Extract the bottom visual signature of image, and different classes of bottom visual signature is carried out being combined to form comprehensive characteristics;
According to described comprehensive characteristics design of graphics as Semantic detection model, and improve described image, semantic detection model by incremental learning method;
One image to be retrieved is provided; And
The image, semantic detection model after improving is utilized to mark all image, semantics contained in described image to be retrieved;
Image, semantic classification according to mark carries out image display.
Preferably, in described image semantic retrieving method, described bottom visual signature comprises color characteristic, shape facility, textural characteristics and local binary feature.
Preferably, in described image semantic retrieving method, the process building described image, semantic detection model according to described comprehensive characteristics comprises:
Described comprehensive characteristics is mapped as the summit of semi-supervised adaptive probability hypergraph model;
The cohesion between each summit is calculated according to the Euclidean distance between each summit;
The threshold function table that one is weighed close and distant relation between summit is constructed according to all intimate angle value;
The super limit belonging to each summit is determined according to described threshold function table.
Preferably, in described image semantic retrieving method, the process being improved described image, semantic detection model by incremental learning method is comprised:
One training set of images is provided,
The training sample choosing a part of unknown mark from described training set of images adds in described image, semantic detection model;
The training sample of robustness to this part unknown mark according to described image, semantic detection model screens;
The mark that the training sample chosen and study obtain is joined in the sample set marked, and reconstructs hypergraph to improve described image, semantic detection model.
Preferably, in described image semantic retrieving method, it is semantic that described training set of images has personage, landscape and still life three class.
Preferably, in described image semantic retrieving method, described image, semantic detection model is utilized to comprise the process that all image, semantics contained in described image to be retrieved mark:
The different semantic probable value of each sample in described image to be retrieved is calculated according to the image, semantic detection model after improving;
Different semantic discriminant function is defined respectively according to the probable value calculated;
If the maximum semantic probable value of sample is greater than this kind of semantic corresponding discriminant score, then this sample is normally marked; Otherwise, be foreign peoples by the semantic tagger of this sample.
The present invention also provides a kind of Semantic Image Retrieval system, and described Semantic Image Retrieval system comprises:
MIM message input module, for obtaining image to be retrieved from system peripherals;
Bottom Visual Feature Retrieval Process module, is connected with described MIM message input module, and extracts the bottom visual signature of described image to be retrieved via described MIM message input module;
Image, semantic detection module, with described bottom Visual Feature Retrieval Process model calling, and marks all image, semantics contained in described image to be retrieved according to the bottom visual signature that described bottom Visual Feature Retrieval Process module is extracted;
Display terminal, is connected with described image, semantic detection module, and carries out image display according to the classification of the image, semantic of described image, semantic detection module mark.
Preferably, in described Semantic Image Retrieval system, described image, semantic detection module is provided with image, semantic detection model, and described image, semantic detection model is the semi-supervised adaptive probability hypergraph model obtained based on incremental learning method.
Preferably, in described Semantic Image Retrieval system, the semantic classes that described image, semantic detection model learnt comprises personage, landscape, still life, and the image, semantic that described image, semantic detection model did not learn is decided to be foreign peoples.
Preferably, in described Semantic Image Retrieval system, described display terminal selects display tone according to the semantic classes of described image to be retrieved.
In image semantic retrieving method provided by the invention and system thereof, the image, semantic detection model after improving is utilized to retrieve in image to be detected all image, semantics contained, comprise the image, semantic that image, semantic detection model did not learn, and give rational semantic tagger.And display terminal can select different displaying schemes according to the semantic classes of image, realizes intelligent display.
Accompanying drawing explanation
Fig. 1 is the process flow diagram of the image semantic retrieving method of the embodiment of the present invention;
Fig. 2 is the structural representation of the Semantic Image Retrieval system of the embodiment of the present invention.
Embodiment
The image semantic retrieving method proposed the present invention below in conjunction with the drawings and specific embodiments and system thereof are described in further detail.According to the following describes and claims, advantages and features of the invention will be clearer.It should be noted that, accompanying drawing all adopts the form that simplifies very much and all uses non-ratio accurately, only in order to object that is convenient, the aid illustration embodiment of the present invention lucidly.
Please refer to Fig. 1, it is the process flow diagram of the image semantic retrieving method of the embodiment of the present invention.As shown in Figure 1, described image semantic retrieving method comprises:
Step S10: the bottom visual signature extracting image, and different classes of bottom visual signature is carried out being combined to form comprehensive characteristics;
Step S20: according to described comprehensive characteristics design of graphics as Semantic detection model, and improve described image, semantic detection model by incremental learning method;
Step S30 a: image to be retrieved is provided;
Step S40: utilize the image, semantic detection model after improving to mark all image, semantics contained in described image to be retrieved;
Step S50: the image, semantic classification according to mark carries out image display.
Concrete, first, extract the bottom visual signature of image.In the present embodiment, described bottom visual signature comprises color characteristic, shape facility, textural characteristics and local this four category feature of binary feature.In other embodiments, described bottom visual signature is color characteristic, shape facility, textural characteristics and local binary feature not necessarily, also can be other features, in this no limit.After the bottom Visual Feature Retrieval Process of image, each category feature is combined, be formed as a kind of comprehensive characteristics.
Then, according to described comprehensive characteristics design of graphics as Semantic detection model.The detailed process building described image, semantic detection model comprises:
Step S21: the summit described comprehensive characteristics being mapped as semi-supervised adaptive probability hypergraph model;
Step S22: calculate the cohesion between each summit according to the Euclidean distance between each summit;
Step S23: construct the threshold function table that is weighed close and distant relation between summit according to all intimate angle value;
Step S24: determine the super limit belonging to each summit according to described threshold function table.
If the intimate angle value on a summit is greater than this threshold value, then judge that this summit belongs to certain super limit.After super limit belonging to each summit is determined, described image, semantic detection model has built.
Afterwards, described image, semantic detection model is improved by incremental learning method.The detailed process that described image, semantic detection model carries out incremental learning comprises:
Step S25 a: training set of images is provided;
Step S26: the training sample choosing a part of unknown mark from described training set of images adds in described image, semantic detection model;
Step S27: the training sample of robustness to this part unknown mark according to described image, semantic detection model screens;
Step S28: the mark that the training sample chosen and study obtain is joined in the sample set marked, and reconstruct hypergraph to improve described image, semantic detection model.
The training sample of known mark and the training sample of a large amount of unmarked semanteme are comprised to the training set of images of described image, semantic detection model training.In incremental learning process, each training sample only choosing a part of unknown mark adds in described image, semantic detection model row filter of going forward side by side, the concrete mode of screening is: first, judges whether the sample added can affect the robustness of described image, semantic detection model; If do not affect robustness, then retain these samples and award corresponding semantic tagger; If affect robustness, then abandon these samples.
Afterwards, an image to be retrieved is provided.Common, in described image to be retrieved, each sample standard deviation has multiple semantic classes.
After this, according to the image, semantic detection model after improving, semantic tagger is carried out to all image, semantics contained in described image to be retrieved, the detailed process that described image to be retrieved carries out semantic tagger is comprised:
Step S41: calculate the different semantic probable value of each sample in described image to be retrieved according to the image, semantic detection model after improving;
Step S42: define different semantic discriminant function respectively according to the probable value calculated;
Step S43: if the maximum semantic probable value of sample is greater than this kind of semantic corresponding discriminant score, then this sample is normally marked; Otherwise, be foreign peoples by the semantic tagger of this sample.
After obtaining the different semantic probable value of each sample in described image to be retrieved by step S41, define the measurement formula (i.e. discriminant function) of all kinds of semanteme respectively with weighting scheme according to these probable values.Thus, every class semanteme has and weighs formula accordingly.If the maximum semantic probable value of a sample is greater than the discriminant score corresponding to this kind of semanteme, then illustrate that this sample packages is containing this kind of semanteme, can normally mark; If the maximum semantic probable value of this sample is less than or equal to the discriminant score corresponding to this kind of semanteme, then illustrating that any class that this sample learnt when not comprising model training is semantic, should be foreign peoples by the semantic tagger of this sample.
In the present embodiment, the all image, semantics contained in described image to be retrieved can be retrieved by the discriminant function of definition, no matter whether these semantemes learnt when model is set up, and can give described image to be retrieved rational semantic tagger with higher accuracy rate.The image semantic retrieving method that the present embodiment provides can realize the multiple semantic retrieval of image, even if the foreign peoples's image, semantic do not learnt when image, semantic detection model is set up also can be detected, retrieval rate is higher.
Finally, image display is carried out according to the image, semantic classification of mark.The image, semantic classification of mark shows tone for image display during foreign peoples does not need to adjust, and when the image, semantic classification of mark is other classes, image display needs to be adjusted to respectively to show tone accordingly.
Accordingly, the present embodiment additionally provides a kind of Semantic Image Retrieval system.
Please refer to Fig. 2, it is the structural representation of the Semantic Image Retrieval system of the embodiment of the present invention.As shown in Figure 2, described Semantic Image Retrieval system 10 comprises: MIM message input module 110, for obtaining image to be retrieved from system peripherals; Bottom Visual Feature Retrieval Process module 120, is connected with described MIM message input module 110, and extracts the bottom visual signature of described image to be retrieved via described MIM message input module 110; Image, semantic detection module 130, is connected with described bottom Visual Feature Retrieval Process module 120, and marks all image, semantics contained in described image to be retrieved according to the bottom visual signature that described bottom Visual Feature Retrieval Process module 120 is extracted; Display terminal 140, is connected with described image, semantic detection module 130, and carries out image display according to the classification of the image, semantic of described image, semantic detection module 130 mark.
Concrete, described MIM message input module 110 obtains image to be retrieved according to user instruction from system peripherals, and described bottom Visual Feature Retrieval Process module 120 extracts the bottom visual signature of described image to be retrieved via described MIM message input module 110.Described bottom visual signature module 120 comprises color characteristic, shape facility, textural characteristics and local binary feature four class from the bottom visual signature of described image zooming-out to be retrieved, and this four classes bottom visual signature is combined into a comprehensive characteristics flows to described image, semantic detection module 130.
Described image, semantic detection module 130 is provided with image, semantic detection model, and described image, semantic detection model is the semi-supervised adaptive probability hypergraph model obtained based on incremental learning.Described image, semantic detection module 130 calculates the different semantic probable value of each sample in described image to be retrieved according to described image, semantic detection model, and define different semantic discriminant functions respectively according to the probable value calculated, retrieve by described discriminant function all image, semantics that described image to be retrieved contains and carry out semantic tagger.
If the maximum semantic probable value of a sample is greater than the discriminant score corresponding to this kind of semanteme, then normally mark; If the maximum semantic probable value of this sample is less than or equal to the discriminant score corresponding to this kind of semanteme, be then foreign peoples by the semantic tagger of this sample.
In the present embodiment, adopt have personage, landscape, still life three class semanteme training set of images to the training of described Semantic Image Retrieval model.Therefore, the semantic classes that described Semantic Image Retrieval model obtains comprises personage, landscape, still life and foreign peoples.In other embodiments, different semantic training set of images can be adopted the training of described Semantic Image Retrieval model, the semantic classes that described Semantic Image Retrieval model obtains can comprise personage, landscape, still life and foreign peoples, also can comprise other semantic classess, in this no limit.
After semantic tagger completes, the semanteme of image is delivered to display terminal 140 by described image, semantic detection module 130, and described display terminal 140 carries out image display according to the semantic classes of image.Wherein, when semantic classes is personage, landscape, still life, described display terminal 140 needs the display tone adjusting image, and when semantic classes is foreign peoples, described display terminal 140 does not need the display tone adjusting image.
In the present embodiment, if the semantic classes of image is personage, then adopts warm tones display, highlight soft; If the semantic classes of image is landscape, then adopts cool tone display, increase contrast, strengthen beautiful scenery to the impulsive force of people; If the semantic classes of image is still life, then adopt inclined warm tones to show, described inclined warm tones, between described warm tones and cool tone, embodies a kind of quiet; If the image of foreign peoples's semanteme, then by tone display originally.
When adopting described Semantic Image Retrieval system 10 to show a certain image, the image of input first obtains image, semantic by described Semantic Image Retrieval model, afterwards by described display terminal 140 be this image select be applicable to displaying scheme, described display terminal 140 is that image selects suitable tone according to the semantic classes of image, can realize more intelligent beautiful figure and show.
To sum up, in the image semantic retrieving method provided in the embodiment of the present invention and system thereof, the image, semantic detection model after improving is utilized to define different semantic discriminant function respectively, thus retrieve in image to be detected all image, semantics contained, comprise the image, semantic that image, semantic detection model did not learn, and all give rational semantic tagger.And, different displaying schemes can be selected according to the semantic classes of image during final display, realize intelligent display.
Foregoing description is only the description to present pre-ferred embodiments, any restriction not to the scope of the invention, and any change that the those of ordinary skill in field of the present invention does according to above-mentioned disclosure, modification, all belong to the protection domain of claims.
Claims (10)
1. an image semantic retrieving method, is characterized in that, comprising:
Extract the bottom visual signature of image, and different classes of bottom visual signature is carried out being combined to form comprehensive characteristics;
According to described comprehensive characteristics design of graphics as Semantic detection model, and improve described image, semantic detection model by incremental learning method;
One image to be retrieved is provided; And
The image, semantic detection model after improving is utilized to mark all image, semantics contained in described image to be retrieved;
Image, semantic classification according to mark carries out image display.
2. image semantic retrieving method as claimed in claim 1, is characterized in that, described bottom visual signature comprises color characteristic, shape facility, textural characteristics and local binary feature.
3. image semantic retrieving method as claimed in claim 1, it is characterized in that, the process building described image, semantic detection model according to described comprehensive characteristics comprises:
Described comprehensive characteristics is mapped as the summit of semi-supervised adaptive probability hypergraph model;
The cohesion between each summit is calculated according to the Euclidean distance between each summit;
The threshold function table that one is weighed close and distant relation between summit is constructed according to all intimate angle value;
The super limit belonging to each summit is determined according to described threshold function table.
4. image semantic retrieving method as claimed in claim 1, it is characterized in that, the process being improved described image, semantic detection model by incremental learning method is comprised:
One training set of images is provided,
The training sample choosing a part of unknown mark from described training set of images adds in described image, semantic detection model;
The training sample of robustness to this part unknown mark according to described image, semantic detection model screens;
The mark that the training sample chosen and study obtain is joined in the sample set marked, and reconstructs hypergraph to improve described image, semantic detection model.
5. image semantic retrieving method as claimed in claim 4, is characterized in that, it is semantic that described training set of images has personage, landscape and still life three class.
6. image semantic retrieving method as claimed in claim 1, is characterized in that, utilize described image, semantic detection model to comprise the process that all image, semantics contained in described image to be retrieved mark:
The different semantic probable value of each sample in described image to be retrieved is calculated according to the image, semantic detection model after improving;
Different semantic discriminant function is defined respectively according to the probable value calculated;
If the maximum semantic probable value of sample is greater than this kind of semantic corresponding discriminant score, then this sample is normally marked; Otherwise, be foreign peoples by the semantic tagger of this sample.
7. a Semantic Image Retrieval system, is characterized in that, comprising:
MIM message input module, for obtaining image to be retrieved from system peripherals;
Bottom Visual Feature Retrieval Process module, is connected with described MIM message input module, and extracts the bottom visual signature of described image to be retrieved via described MIM message input module;
Image, semantic detection module, with described bottom Visual Feature Retrieval Process model calling, and marks all image, semantics contained in described image to be retrieved according to the bottom visual signature that described bottom Visual Feature Retrieval Process module is extracted;
Display terminal, is connected with described image, semantic detection module, and carries out image display according to the classification of the image, semantic of described image, semantic detection module mark.
8. Semantic Image Retrieval system as claimed in claim 7, it is characterized in that, described image, semantic detection module is provided with image, semantic detection model, and described image, semantic detection model is the semi-supervised adaptive probability hypergraph model obtained based on incremental learning method.
9. Semantic Image Retrieval system as claimed in claim 8, it is characterized in that, the semantic classes that described image, semantic detection model learnt comprises personage, landscape, still life, and the image, semantic that described image, semantic detection model did not learn is decided to be foreign peoples.
10. Semantic Image Retrieval system as claimed in claim 7, is characterized in that, described display terminal selects display tone according to the semantic classes of described image to be retrieved.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510306112.4A CN104834757A (en) | 2015-06-05 | 2015-06-05 | Image semantic retrieval method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510306112.4A CN104834757A (en) | 2015-06-05 | 2015-06-05 | Image semantic retrieval method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104834757A true CN104834757A (en) | 2015-08-12 |
Family
ID=53812643
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510306112.4A Pending CN104834757A (en) | 2015-06-05 | 2015-06-05 | Image semantic retrieval method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104834757A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105740888A (en) * | 2016-01-26 | 2016-07-06 | 天津大学 | Joint embedded model for zero sample learning |
| CN105740879A (en) * | 2016-01-15 | 2016-07-06 | 天津大学 | Zero-sample image classification method based on multi-mode discriminant analysis |
| CN106973244A (en) * | 2016-01-13 | 2017-07-21 | 奥多比公司 | Using it is Weakly supervised for image match somebody with somebody captions |
| CN111127398A (en) * | 2019-11-25 | 2020-05-08 | 深圳前海微众银行股份有限公司 | Method and device for detecting certificate photo duplication, terminal equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030123737A1 (en) * | 2001-12-27 | 2003-07-03 | Aleksandra Mojsilovic | Perceptual method for browsing, searching, querying and visualizing collections of digital images |
| CN102073748A (en) * | 2011-03-08 | 2011-05-25 | 武汉大学 | Visual keyword based remote sensing image semantic searching method |
| CN103390063A (en) * | 2013-07-31 | 2013-11-13 | 南京大学 | Search method for relevance feedback images based on ant colony algorithm and probability hypergraph |
| CN103678483A (en) * | 2013-10-24 | 2014-03-26 | 江苏大学 | Video semantic analysis method based on self-adaption probability hypergraph and semi-supervised learning |
-
2015
- 2015-06-05 CN CN201510306112.4A patent/CN104834757A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030123737A1 (en) * | 2001-12-27 | 2003-07-03 | Aleksandra Mojsilovic | Perceptual method for browsing, searching, querying and visualizing collections of digital images |
| CN102073748A (en) * | 2011-03-08 | 2011-05-25 | 武汉大学 | Visual keyword based remote sensing image semantic searching method |
| CN103390063A (en) * | 2013-07-31 | 2013-11-13 | 南京大学 | Search method for relevance feedback images based on ant colony algorithm and probability hypergraph |
| CN103678483A (en) * | 2013-10-24 | 2014-03-26 | 江苏大学 | Video semantic analysis method based on self-adaption probability hypergraph and semi-supervised learning |
Non-Patent Citations (1)
| Title |
|---|
| 潘刚: "基于语义的图像显示适配技术研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106973244A (en) * | 2016-01-13 | 2017-07-21 | 奥多比公司 | Using it is Weakly supervised for image match somebody with somebody captions |
| CN106973244B (en) * | 2016-01-13 | 2021-04-20 | 奥多比公司 | Method and system for automatically generating image captions using weak supervision data |
| CN105740879A (en) * | 2016-01-15 | 2016-07-06 | 天津大学 | Zero-sample image classification method based on multi-mode discriminant analysis |
| CN105740879B (en) * | 2016-01-15 | 2019-05-21 | 天津大学 | The zero sample image classification method based on multi-modal discriminant analysis |
| CN105740888A (en) * | 2016-01-26 | 2016-07-06 | 天津大学 | Joint embedded model for zero sample learning |
| CN111127398A (en) * | 2019-11-25 | 2020-05-08 | 深圳前海微众银行股份有限公司 | Method and device for detecting certificate photo duplication, terminal equipment and storage medium |
| CN111127398B (en) * | 2019-11-25 | 2023-09-26 | 深圳前海微众银行股份有限公司 | Certificate photo copying detection method and device, terminal equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104268603B (en) | Intelligent marking method and system for text objective questions | |
| He et al. | Improving the normalized difference built-up index to map urban built-up areas using a semiautomatic segmentation approach | |
| CN102968637B (en) | Complicated background image and character division method | |
| CN105005786B (en) | A kind of texture image classification method based on BoF and multiple features fusion | |
| CN103049446B (en) | A kind of image search method and device | |
| US10163227B1 (en) | Image file compression using dummy data for non-salient portions of images | |
| US20080025647A1 (en) | Image management methods, image management systems, and articles of manufacture | |
| US9652561B2 (en) | Method, system and program for processing input data to facilitate selection of a corresponding tag candidate | |
| CN104142961B (en) | The logic processing device of composite diagram and logical process method in format document | |
| CN109165350A (en) | A kind of information recommendation method and system based on deep knowledge perception | |
| CN104834757A (en) | Image semantic retrieval method and system | |
| CN104142995B (en) | The social event recognition methods of view-based access control model attribute | |
| CN102385592B (en) | Image concept detection method and device | |
| CN105809146A (en) | Image scene recognition method and device | |
| CN108734210A (en) | A kind of method for checking object based on cross-module state multi-scale feature fusion | |
| CN108132887A (en) | User interface method of calibration, device, software testing system, terminal and medium | |
| CN110008309A (en) | A kind of short phrase picking method and device | |
| CN105260428A (en) | Picture processing method and apparatus | |
| CN105718555A (en) | Hierarchical semantic description based image retrieving method | |
| CN101388020A (en) | Composite image search method based on content | |
| CN109308324A (en) | A kind of image search method and system based on hand drawing style recommendation | |
| CN1973299A (en) | Image search method and image search device | |
| CN110598018A (en) | Sketch image retrieval method based on cooperative attention | |
| CN109271546A (en) | The foundation of image retrieval Feature Selection Model, Database and search method | |
| CN113223042A (en) | Intelligent acquisition method and equipment for remote sensing image deep learning sample |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20150812 |