Content-based image retrieval technology has been a hot spot for a long time, but the effect of retrieval based on a single bottom feature is very poor. The paper brought up a method based on multi-feature for orthodontics images, doctors could find out the patients’ diagnosis information according to some similar images. The basic theory is to extract color feature by combining the cumulative histograms and spatial information, achieving image segmentation by combining Canny operator and Ostu threshold method, and then extracting seven rectangular-shape features. Finally it carries out image retrieval based on multi-feature by combining color component and shape component. The results of tests show that this method has a better effect for orthodontics images.