Endoscope Image Retrieval Based on Color Feature Fusion

In view of endoscope images´ color distribution, a non-even color quantization method in HSV space is proposed in this paper. Then based on this quantization many kinds of color feature including the color histogram, the main color, the mean color, the accumulative histogram, the color pair, the dominant color of partition, the color correlogram and color moments are extracted. Comparing the retrieval efficiency of each feature, a variety of colors features are combined for synthesis feature retrieval. Finally, a relevance feedback technology is joined in synthesis feature retrieval. Experiments have proved that combining the color correlogram with histogram joining relevance feedback retrieval can get better efficiency.