Technology Research, Content-based Image Retrieval

Extracting features from image effectively is the key issues in Content-Based Image Retrieval.In This paper, we analyze some methods of extracting the color, texture features which belong to the low-layer image features from image content. In the research of extracting color features, this dissertation mainly discuss and analyze three kinds of arithmetic based on color histogram in HSV color space.Because the ability of expressing image information with color feature is limited, texture feature is presented. The method of extracting texture feature is adopted, which are based on Gray Level Co-occurrence Matrix. This dissertation applies a simplified calculation on Gray Level Co-occurrence Matrix to image retrieval, which reduce the calculation of feature vector, improve the efficiency of the image retrieval. The performance of this experiment is similar to the traditional method, so it is feasible that apply the simplified calculation on Gray Level Co-occurrence Matrix to the image retrieval.In this dissertation a method using combined color and texture feature is also given. In the plenty of experiment results compared with which using single feature, the experiments give good results. So combine with multi image feature can describe image content generally and exactly.Finally, a prototype image retrieval system is constructed, which realizes the content-based image retrieval. The experiment with the system proves the validity of the proposed methods.