An image retrieval algorithm based on color and spatial distribution of salient points was proposed. Firstly, Harris color point detector was used to extract color salient points. Secondly, salient points were regarded as clues, and a circular color histogram based on salient points was designed, which combined local color feature with spatial distribution structure of salient points effectively. Finally, information entropy was used, which considered spatial distribution information and decreased complexity. The experimental results show that this method is easily manipulated, and has better performance than image retrieval based on global characteristics.