A new image retrieval algorithm based on image entropy and spatial distribution entropy was presented. At first a more robust method, which can remove the influence of the symmetry of entropy, was proposed to extract the global color feature. Then color spatial distribution entropy vector for each color channel was also introduced to represent the spatial color information. After that, the moments were adopted to reduce the dimension of color spatial distribution entropy. In the end, a low dimensional vector which includes the global and spatial information was used as index for color image retrieval. The experiment results show that the new method gives better performance than color histogram.