CN102831187A

CN102831187A - Content-based image retrieval system

Info

Publication number: CN102831187A
Application number: CN2012102732202A
Authority: CN
Inventors: 吴军
Original assignee: CHENGDU ZHONGHE YUNSHENG TECHNOLOGY CO LTD
Current assignee: CHENGDU ZHONGHE YUNSHENG TECHNOLOGY CO LTD
Priority date: 2012-08-02
Filing date: 2012-08-02
Publication date: 2012-12-19

Abstract

The invention relates to a content-based image retrieval system. An output end of a preprocessing module is respectively connected with a target identification module and a feature extracting module; an output end of the target identification module is connected with an input end of the feature extracting module; the target identification module and the feature extracting module are respectively connected with a knowledge aiding module in two ways; a query interface module is connected with a retrieval engine module in two ways; the retrieval engine module is connected with a retrieval filter module in two ways; the query interface module and the retrieval engine module are simultaneously connected with the knowledge aiding module in two ways; and output ends of the feature extracting module and the retrieval filter module are connected with a database. The content-based image retrieval system can be frequently interacted with a user through a visual interface, so that the user can conveniently set query, estimate retrieval results and improve the retrieval results.

Description

The CBIR system

Technical field

The present invention relates to the field of image retrieval technologies, especially a kind of CBIR system.

Background technology

In recent years, along with the develop rapidly of multimedia technology and computer network, the capacity of global digital picture increases just with surprising rapidity.No matter be military or civil equipment, all can the generation capacity be equivalent to the image of thousands of megabyte every day.A large amount of Useful Informations have been comprised in these digital pictures.Yet because these images are to be distributed in disorderly all over the world, the information that comprises in the image can't be visited and utilized effectively.This just require a kind of can be fast and search the technology of access images, just so-called image retrieval technologies exactly.Since the seventies in 20th century, under data base set was unified the common promotion of computer vision two big research fields, image retrieval technologies became a very active research field gradually.Database is to study image retrieval technologies from different angles with computer vision two big fields, the former text based, and the latter is based on vision.

The history of text-based image retrieval technology (text-based image retrieval) can be traced back to phase late 1970s.Popular at that time image indexing system be with image as an object of storing in the database, described with key word or free text.The textual description that query manipulation is based on this image accurately matees or probability match, and the retrieval model of some system still has dictionary to support.In addition, technology such as view data model, multi-dimensional indexing, query evaluation all grow up under such framework.Yet the text-based image retrieval technology exists serious problem fully.At first, present computer vision and artificial intelligence technology all can't mark image automatically, and must depend on manual work image are made mark.This work is not only wasted time and energy, and manual mark is inaccurate often or incomplete, also has subjective deviation inevitably.That is to say that different people has different understanding methods to same width of cloth image, the difference of this subjective understanding will cause the mismatch error in the image retrieval.In addition, the abundant visual signature (color or texture etc.) that is comprised in the image often can't be described with text objectively.

At the initial stage nineties, along with the appearance of large scale digital image library, above-mentioned problem becomes more and more sharp-pointed.For overcoming these problems, CBIR technology (content-based image retrieval) is arisen at the historic moment.Be different from the original system image is carried out the artificial way that marks, the vision content characteristic that the content-based retrieval technology is extracted every width of cloth image automatically is as its index, like color, texture, shape etc.After this in several years, the many technical developments in this research field are got up, and large quantities of research property or commercial image indexing system is established.The progress of computer vision technique is mainly given the credit in the development in this field, and the detailed introduction to this field is arranged in document

Summary of the invention

The technical matters that the present invention will solve is: in order to overcome the problem that exists in above-mentioned, a kind of CBIR system is provided, its project organization rationally and can make things convenient for user inquiring and assessment retrieval.

The technical solution adopted for the present invention to solve the technical problems is: a kind of CBIR system; Comprise feature extraction subsystem and inquiry subsystem; Described feature extraction subsystem comprises pre-processing module and target identification module and the characteristic extracting module that is connected with the view data output terminal; The pre-processing module output terminal is connected with characteristic extracting module with the target identification module respectively, and target identification module output terminal is connected with the characteristic extracting module input end, the two-way respectively knowledge supplementary module that is connected with of target identification module and characteristic extracting module; Pre-processing module comprises the conversion of picture format, the unification of size; Functions such as the enhancing of image and denoising, for the feature extraction of image lays the first stone, the target identification module provides a kind of instrument for the user; With user's interest zone or destination object in the mode identification image of full-automatic or semi-automatic (needing user intervention), so that carry out feature extraction and inquiry to target.When carrying out the entirety retrieval, utilize global characteristics, at this moment without the target identification function.Target identification is optional, and characteristic extracting module is carried out feature extraction to image data base, extracts user's interest, is fit to the characteristic that retrieval requires.Feature extraction can be of overall importance, i.e. entire image also can be to certain target, i.e. subregion in the image is like people's face etc.;

Described inquiry subsystem comprises and the two-way query interface module that is connected of user and search engine module and index filtering module; The query interface module is connected with the search engine module is two-way; The search engine module is connected with the index filtering module is two-way; The query interface module is connected with the knowledge supplementary module is two-way with the search engine module simultaneously; Retrieval is to utilize the distance function between the characteristic to carry out similarity retrieval, and search engine reaches the purpose of quick search through index/filtering module, thereby can be applied in the large database.Filtrator acts on total data, and the data acquisition that filters out matees with high dimensional feature again to be retrieved.Index is used for low dimensional feature, can set index to be connected with database to accelerate described characteristic extracting module of retrieval and index filtering module output terminal with R.

Described database is made up of image library and feature database and knowledge base, and image library is digitized image information, and the characteristic that feature database comprises user's input and pre-service be the content characteristic of extraction automatically.Knowledge base comprises special and world knowledge, helps query optimization and matees fast, and knowledge representation can be changed to be suitable for various application in the knowledge base.

The invention has the beneficial effects as follows that CBIR of the present invention system carries out frequent alternately through visualization interface and user, be convenient to the user and can conveniently construct inquiry, assessment result for retrieval and improvement result for retrieval.

Description of drawings

Below in conjunction with accompanying drawing and embodiment the present invention is further specified.

Fig. 1 is a structured flowchart of the present invention.

11. pre-processing module among the figure, 12. target identification modules, 13. characteristic extracting module, 21. query interface modules, 22. search engine modules, 23. index filtering modules, 3. knowledge supplementary module, 4. database, 41. image libraries, 42. feature databases, 43. knowledge bases.

Embodiment

Combine accompanying drawing that the present invention is done further detailed explanation now.These accompanying drawings are the synoptic diagram of simplification, basic structure of the present invention only is described in a schematic way, so it only show the formation relevant with the present invention.

CBIR system as shown in Figure 1; Comprise feature extraction subsystem and inquiry subsystem; The feature extraction subsystem comprises the pre-processing module 11 and target identification module 12 and characteristic extracting module 13 that is connected with the view data output terminal; Pre-processing module 11 output terminals are connected with characteristic extracting module 13 with target identification module 12 respectively; Target identification module 12 output terminals are connected with characteristic extracting module 13 input ends; Target identification module 12 and the two-way respectively knowledge supplementary module 3 that is connected with of characteristic extracting module 13; The inquiry subsystem comprises and the two-way query interface module that is connected 21 of user and search engine module 22 and index filtering module 23, query interface module 21 and 23 two-way connections of search engine module, search engine module 22 and 23 two-way connections of index filtering module; Query interface module 21 and search engine module 22 simultaneously and 3 two-way connections of knowledge supplementary module, characteristic extracting module 13 and index filtering module 23 output terminals are connected with the database of being made up of image library 41 and feature database 42 and knowledge base 43 4.

With above-mentioned foundation desirable embodiment of the present invention is enlightenment, and through above-mentioned description, the related work personnel can carry out various change and modification fully in the scope that does not depart from this invention technological thought.The technical scope of this invention is not limited to the content on the instructions, must confirm its technical scope according to the claim scope.

Claims

1. CBIR system; It is characterized in that: comprise feature extraction subsystem and inquiry subsystem; Described feature extraction subsystem comprises pre-processing module (1) and target identification module (12) and the characteristic extracting module (13) that is connected with the view data output terminal; Pre-processing module (1) output terminal is connected with characteristic extracting module (13) with target identification module (12) respectively; Target identification module (12) output terminal is connected with characteristic extracting module (13) input end; The two-way respectively knowledge supplementary module (3) that is connected with of target identification module (12) and characteristic extracting module (13); Described inquiry subsystem comprises and the two-way query interface module (21) that is connected of user and search engine module (22) and index filtering module (23); Query interface module (21) and two-way connection of search engine module (23); Search engine module (22) and two-way connection of index filtering module (23), query interface module (21) and search engine module (22) while and two-way connection of knowledge supplementary module (3), described characteristic extracting module (13) and index filtering module (23) output terminal are connected with database (4).

2. CBIR according to claim 1 system is characterized in that: described database (4) is made up of image library (41) and feature database (42) and knowledge base (43).