Image quality assessment based on local variance and structure similarity

Abstract

To overcome the unequal sensitivity to different distortion of structural similarity,(SSIM) a new image quality assessment method based on local variance and SSIM is presented.The comparison of the local variance distribution of two images is better suited to assess the detail structure similarity,which is more sensitive to some distortion such as blurring than SSIM but less sensitive to some distortion such as noise than SSIM.The assessment method proposed in this paper is a combination of local variance distribution and SSIM,which balances their different sensitivity to different image distortions and gives rational results that are consistent with visual quality.The experimental result shows that the method is better consistent with human visual characteristics than SSIM.