Font Size: a A A

The Research Of Content-Based Image Retrieval Technique And System Implementation

Posted on:2008-10-14Degree:MasterType:Thesis
Country:ChinaCandidate:L X ZhuFull Text:PDF
GTID:2178360215493554Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Content-Based Image Retrieval (CBIR) is a kind of technique basedon the image content, such as people's comprehension and understanding,to retrieve images by weighing the similarity of two images. And thepurpose of CBIR is to organize, manage and search huge image dataeffectively. With the development and popularization of digital technology,computer technology and network technology, more and more imageinformation has arisen, which makes CBIR an active research area in recentyears.This paper contains the research of CBIR and prototype systemimplementation. The concrete content is as follows:1. In this paper, the background, application and research actuality ofCBIR and some distinctive CBIR systems are introduced. And the keytechnologies of CBIR are discussed thoroughly.2. Based on the further study of color feature-derived and similaritymatching, a new algorithm of references color set generalized average histogram is brought forward, which improves the traditional colorquantifier and similarity matching arithmetic.3. In order to lessen the semantic gap between the low-level visualfeatures and the high-level semantic features, a novel dynamic relevancefeedback algorithm is put forward, in which positive and negativefeedbacks are used to adjust region weights in Image Retrieval. And theexperimental results demonstrated that this algorithm achieves betterperformance.4. To meet the demand of better description of image semantic andhuman conception, a similarity matching method based on fuzzy logic isdesigned, referencing on fuzzy inference on CBIR. The experimentalresults show that this algorithm achieves satisfying robustness, validity andfeasibility.5. A CBIR prototype system is presented here, and the architecture,function modules, system flows and interfaces of this system are describedin detail. In the end of this paper, the main work of our research is reviewed,the main problems and techniques in this field are analyzed and the futuredirection of development is suggested.
Keywords/Search Tags:Content-Based Image Retrieval, generalized image, relevance feedback, fuzzy inference, fuzzy density
PDF Full Text Request
Related items