Image Retrieval Based On Color Features And Realization

In recent years, with the rapid developing of the Internet, on-line multimedia information also increased rapid. Faced with the large number of informations, people eager to be able to quickly and efficiently search the required multimedia information. Effective organization, management and retrieval to large image data has become an urgent problem to be solved.Content-based image retrieval technology is a research hot point in many research institutions abroad, a lot of reaserchers are engaged to the reaserche of the subject and get some research results. In this thesis, I read a lot of reference about CBIR (Content Based Image Retrieval) and reaserch the similarity matching algorithm, and analysis these algorithm's effect.In content-based image retrieval, the image color is an important feature of vision, it has been widely used. This thesis concentrated on three key issues of color image retrieval : the reprention of colors, the similarity measure and feature extraction method This thesis researched the tranditional color-based feature extraction methods, discussed the advantage and the disadvantage of existing feature extraction methods on considering the color distribution and spatial location. In the prototype system, this thesis implemented these tranditional algorithm, then contrive two improved feature extraction method: a unequal partition method and a automatic popedom decision method.This thesis used Visual C ++ 6.0 as a development tool, SQL Server 2000 as database to developed a simple color-based image retrieval system prototype. Then used a lots of image to test the prototype and the improved algorithm, compared the results in chart and make some conclusion.In this thesis, the researching and practice of the promotion of content-based image database retrieval technology has a certain reference value and practical significance.