A Novel Full-Reference Color Image Quality Assessment Based on Energy Computation in the Wavelet Domain

This paper presents a novel full reference (FR) image quality assessment algorithm by depicting the subband characteristics in the wavelet domain. The proposed image quality assesment method is based on the energy estimation in wavelet transformed image. Image quality assessment is achieved by applying a multi-level wavelet decomposition on the both original and enhanced image. Next, the wavelet energy is computed in order to obtain the percentage of energy corrosponding to approximation and vector containing percentage of energy corrosponding to details. Further, the approximate and detail energy levels of both original and enhanced images are compared in order to formulate an image quality assesment. Numerous experiments are conducted on a dozen of image enhancement algorithms. The results presented show that the image with poor contrast in foreground then the background has continuous regular coefficient values. The Probability Density Function (PDF) for such image has relatively lower wavelet energy and skewness compared with background. The proposed scheme not only evaluates global information of an image but also estimates fine detail changes in an enhanced image. Thus, the proposed metric serves as a objective and effective assessment full reference criterion for color image quality assessment. Experimental results presented confirm that the proposed wavelet energy metric is an efficient and useful metric for evaluating the quality of the color image enhancement.