Multi Resolution Enhanced Orthogonal Polynomilas Based Autocorrelogram for Color Image Retrieval

This paper proposes a simple color image retrieval method based on multi resolution enhanced orthogonal polynomials model. In the proposed method, a set of orthogonal polynomials has been chosen and the model coefficients are reordered into multiresolution subband like structure. The subband coefficients are quantized into three levels. Then the weighted autocorrelogram is computed from the quantized subbands in R, G and B color space. The obtained weighted autocoorelogram is termed as global color feature vector and are normalized with Z score normalization. This feature vector is used for retrieving similar images with weighted Manhattan distance metric. The efficiency of the proposed method is experimented on a subset of standard COREL database and the results are compared with existing techniques. The proposed method yields significant retrieval results.