Abstract: First analyzed in this paper are problems of image retrieval methods based on color comparison. And then a new approach to color based image retrieval is presented. It quantizes colors of each image separately, sorts the color histogram, and then compares the difference of corresponding colors ranked by their presence ratio in two images and the difference of colors which contrast strongly with each other. The storage requirement of the image feature is very low and the algorithm works fast. Experimental results are given to show its ability of identifying images and retrieving similar images. Its performance is better than that of the histogram distance method.