The X-ray chest image retrieval based on feature fusion

Based on the analysis of methods of CBIR and chest image characteristic, in this paper, color correlogam, dominant color of partition, gray level co-occurrence matrix, gray-gradient co-occurrence matrix and shape invariant moments were extracted as retrieval feature. After comparison of their retrievals, feature fusion and relevance feedback is proposed. Experiments proved that the combining color, texture with shape feature gets effective retrieval and relevance feedback further more improves retrieval.