Selecting the proper window for SSIM

摘要

The structural similarity index (SSIM) is a well-known metric in the field of image quality assessment (IQA). It is a full-reference metric that uses a sliding window to determine local quality/distortion measures between two images (based on a combination of luminance, contrast, and structural measurements from each image's window), and combines the results to obtain a single value indicating the quality of the distorted image relative to its perfect reference image. The metric uses a window of arbitrary size and shape to compute the localized measurements, but no work has been done to determine the effects of different window selections. This paper provides insight into these effects by testing various window modifications against six publicly-available image quality databases. Optimal window results are presented as well as a visual mapping of window parameters vs. performance.