A Novel Structural Similarity with High Distinguishability

摘要

Objective assessment of image quality is important for a number of image processing applications. Compare to the other existing algorithms, the greatest advantage of the structural similarity (SSIM) metric is that the algorithm is based on the structural distortion of image and highly matches human subjectivity. By deeply studying the SSIM, we find it difficult to locate the detailed part of an image distortion clearly. In this paper, we present a novel approach which is called High-Distinguishability Structure Similarity (HD-SSIM) to promote the resolution of the SSIM map. Experiment results show that HD-SSIM is more consistent with HVS than SSIM especially for the images with texture distortion.