Abstract: Content-based image retrieval has broad application prospects, but search efficiency is not high.Based on the advantages of multi-feature fusion and interest points, proposed an Image retrieval algorithm.using Harris algorithm extract interest points, determine Interest Region;then using color histogram, co-occurrence matrix and shape invariant moment extract the color, texture and shape features of interest region;finally using distance function to measure the similarity between images to retrieve the image.Experiments show that this algorithm effectively improves the image retrieval efficiency.