Image Retrieval Using Color Moment Invariant

This paper describes a new color image retrieval method using color moment invariant. The representative colors are computed from each image instead of being fixed in a given color space, thus allowing the feature representation to be accurate as well as compact. Different from previous methods, the proposed method uses small image descriptors and it is adaptive to the context of the image itself by a two-stage clustering technique. Experiments on the image library COIL-100 demonstrate the effectiveness of this method.