Research On Image Feature Extraction Based On Color And Texture And Retrieval Approach

Aimed at the increasing application demands of image information retrieval in each field, the algorithms in color, texture feature, and relevance feedback are explored further based on analyzing the present technique about Content-based Image Retrieval and the confronting problem. The main innovations in the dissertation are:(1)Aimed at the defect that global color feature can't catch the space information of the image, and the problem that the color at the quantification critical edge which quantification area does it should belong to in the algorithm of color histograms, the part lifting scheme algorithm is proposed, which combines lifting scheme and sub regions together.(2)Aimed at the failure of the present texture feature extraction algorithms, the lifting scheme algorithm is used to extract texture feature. The accuracy of texture description is improved.(3) In order to improve the extracting performance, the relevance feedback algorithm is researched in depth. A feedback algorithm blending relevance Genetic Algorithm is proposed. This algorithm uses positive and negative feedback images'information to adjust feature vector weight, region weight and color-texture weight, and combines with memory and taboo search algorithm, applies the Genetic Algorithm to renew and optimize the example feature vector continuously to catch user's purposes in a short time. According to the algorithms above, a Content-based Image Retrieval system is exploited. And the algorithms'effectiveness is tested and verified. The experimental results show that the system is effective and available.At last, the summary is made, the main results are given, and the research work in the future is projected.