Image Retrieval Based on Mutual Information

In order to solve the disadvantage of ignoring the visual feature in image retrieval, a novel method based on mutual information was presented incorporating the different sensitivity along variant directions of human visual system. Firstly, combining the visual feature with principle of the block truncation code (BTC), the significant image feature is extracted and a novel descriptor is defined. Based on the analysis of the characters of the descriptors, the mutual information is chosen to image retrieval. Compared with the other algorithms, experiments show the proposed approach has good performance.