The aim of this paper is to increase the success rate of CBIR system with low computational complexity. The success rate of CBIR system depends on localization of the image to be retrieved. This can be achieved by using textons of R, G, B planes of the image which describes the shape. This paper proposes 3 × 3 grids to extract the textons with low computational complexity. The proposed method is based on the texels (low level features) of textons extracted from R,G,B channels of an image as chromatic changes also give shape information. The proposed method is tested on Corel database with more than 1000 natural images. The results demonstrate that it is more efficient than texton co-occurrence matrix, texton multi histogram methods. It has good discrimination power of color, texture and shape features when compared to that of TCM and TMH methods.