Application And Research Of Content-based Image Retrieval

With the rapid development of Internet and digital equipments,the amount of multimedia information being produced,there is lots of opportunity to get the enriched multimedia information.In order to quickly and accurately find the images which have rich content from large amount of images,the Content-based image retrieval(CBIR)technology has been studied a lot currently.Most of traditional image retrieval only provides the retrieval method based on the image's key words,which image is marked by people firstly and is searched to find the retrieval image.This method is simple and the result of search is accurate, but it have to spend lots of time to mark these key words and there is the "semantic gap" problem that cannot express image content precisely.Therefore,the content-based image retrieval technology is proposed by researchers since the early 1990's.CBIR actually is the kind of Fuzzy inquiry technology.CBIR can use the visual features,such as color,texture,shape and space relationship,to find the waited image.The content-based image retrieval technology is deeply researched in this paper. The construct of CBIR and retrieval standards are analysized after introducing the basic CBIR's background,application fields and current research situation both at home and abroad.The concepts and characteristics of the feature extraction,similarity metric and relevant feedback are proposed by studying the structure of CBIR.In the part of the feature extraction,the definitions and math expressions of color,texture,shape and space relationship features are described in detail,The characteristic method and usually express way of similarity metric are introduced. What's more,the relevant feedback technology in the CBIR is also studied.In this paper,a simple cotent-based image retrieval system is proposed,which is based on the color feature of color moment and the texture feature of gray level co-occurance matrix by connectting the principle of content-based image retrieval. This system is designed by Visual C++ 6.0 and SQL Server2000 database technology. The color moment value and gray level co-occurance matrix value of the input example image are computed,and compare with the image feature set by using algorithm of similarity metric,at last the result of similarity metric is disposed on the screen.After analysizing system's each module function,the result of the experiment is shown that the retrieval algorithm we used has the better retrieval efficiency and accuracy,and got better results.