Image Difference Estimation and Noise Removing

PSNR (Peak Signal to Noise Ratio) metric is typically used to estimate difference between two given images. Recent time more reliable metric (SSIM – Structural Similarity) was introduced to detect image changes/noise/etc. This article introduces more advanced approach for the image distance measurement, based on weighted sum of the image structure estimation and edges detection. Proposed metric (ESSIM – Edgeted Structural Similarity) is more close to the human image difference perception, and provide strong emphasize to pixels near edges. Paper illustrates two image modification cases, where PSNR and SSIM cannot find difference between images, but proposed metric can do it. Also this paper introduces a novel image noise removal approach, based on the weighted mixture of bilateral method and advanced edge detection approach.