Image Quality Assessment Based on Nonsubsampled Contourlet Transform and Structural Similarity

This paper presents a new image quality assessment which is based on structural similarity in nonsubsampled contourlet transform (NSCT) domain. The nonsubsampled contourlet transform is introduced for its ability to represent the images at different scales and directions. Firstly, images were decomposed into subbands with different scales and directions by (NSCT). Secondly, the correlativity indexes between the reference sequences and the comparative sequences respectively were calculated in each subbands. According to information content weighting, the quality of the whole image can be obtained. Experimental results show that the proposed method improves accuracy and robustness of image quality prediction.