A Majorization-Minimization Approach for Edge-Aware Image Smoothing

Abstract

Edge-aware image smoothing plays an important part in practical applications of both computer vision and graphics. In this paper, we propose a majorization-minimization (MM) approach for edge-preserving image smoothing based on the edge-stopping function. With a given edge-stopping function, our theoretical results show that the differences between neighboring pixels in homogeneous regions of the output smoothing image are smaller than a small threshold, while they are larger than a large threshold at edges. We give the process of the MM algorithm solving the image smoothing problem in detail and discuss the convergence of the MM algorithm. Several experiments such as artifact removal, detail enhancement, texture smoothing and tone mapping are given to demonstrate the effectiveness of the proposed approach.