Technology Research, Content-based Image Retrieval

With the rapid development of database system and computer vision, Content-Based Image Retrieval (CBIR) technique has become a very hot point of study. As for this study area, in this paper, we firstly introduce the history and development of CBIR and also the typical CBIR system in foreign country, and summarize briefly on the study of CBIR in nation. Then we introduce the framework of the CBIR system and the function of each module, and also the key technique in CBIR, such' as the image feature extraction and representation, feature—based similarity computation, the standard of performance estimate and the relevance feedback technique .We mainly focus on research and study the technique of CBIR based on color and texture feature, and we manage to conclude and summarize algorithms for each feature, and then select several color and texture extraction algorithms to carry out an experiment. As for color-based image retrieval, we have compared the experimental results with histogram in six color space. The experimental results show that the best retrieval results is using HSV color space, manifested the HSV color space to be consistent with human' s visual sensation. As for texture feature retrieval, we introduce typical co-occurrence matrix algorithms and gray texture matrix algorithms. In texture feature retrieval experimental results, gray texture matrix got the best retrieval results. In addition, we have analyzed the characteristic Gauss normalization experimental result, which indicates the process Gauss normalization retrieval effect distinct enhancement. Finally, we have constructed a simple retrieval system combining color and texture feature. The algorithm we using has considered the color vector and the texture vector weight relations. The experimental results show that the combing feature retrieval has the better retrieval performance than the signal feature retrieval.