Edge-based suboptimal Kuan filtering for speckle noise reduction

Abstract

This paper addresses the Kuan filter designed for speckle noise reduction. The Kuan filter relies on the effective number of looks as a main parameter which is usually adjusted for each image, after several filter runs, in order to tune the filter for best performance. The proposed solution automatically estimates the optimal filter parameter value based on edge pixels statistics, which results in near-optimal performance where the PSNR loss, compared to the best possible filter output, does not exceed 0.1 dB most of the time.