Skip to Main Content
In order to solve the problem of millions image retrieval in science and technology resources database, the authors firstly studied extraction technologies of bottom image features in CBIR. Sub-block histogram method was used in the color feature extraction, and the Gabor wavelet transformation method were used in the texture feature extraction, while the invariant moment method was used in the shape feature extraction, then the authors combined different features for image retrieval, and used relevance feedback based on weight adjustment techniques to make retrieval results more according with user needs. On the basis of the study, the authors designed and implemented Web-based image retrieval experimental system of science and technology resources database. The experiments indicate that feature combination and relevance feedback can improve the image retrieval precision rate of science and technology resources effectively.