Image Quality Assessment Based on DCT and Structural Similarity

According to the property that the human visual system has different sensitivities in the frequency domain, a novel image quality assessment algorithm based on Structural Similarity (SSIM) and DCT is proposed. The method takes frequency information as a part of structural information and gets the frequency structural comparison function by weighting the frequency components based on the importance of sensitivities of human eyes. Then the luminance comparison, contrast comparison, structural comparison as defined in SSIM and frequency structural comparison are combined to yield an over all similarity measure. Experimental results illustrate that this algorithm has a good consistency with the subjective assessment of human beings for JPEG2000,Gaussian blur, White Noise images and all images in LIVE database, thus it can be used to describe the visual perception of image more reasonable than SSIM,MS-SSIM and PSNR.