Abstract: Color is a kind of basic element composed of image content,and it is also one of the important perceptive features for human to recognize image.A color based image retrieval method is presented in the paper,with emphasis on the choice of color space,the algorithm of color cluster,the distance of color histograms,and the color histograms of image segmentation.The primary goal of image database is to retrieve image data efficiently and accurately,but the efficiency and accuracy of retrieval algorithm are mutually effected.In the process of image retrieval,the multi step similar comparison of coarse granularity and fine granularity is adopted.At the beginning,the coarse granularity comparison is used to filter some irrelevant images,reducing the extent of retrieval,and then the fine granularity comparison is adopted to increase the accuracy of retrieval.