On the convexification of the Perona–Malik diffusion model

Of the image denoising models, the Perona–Malik one has captured the interest of most scholars for its ability to recover semantically important image features. However, this model emanates from an energy functional that is not entirely convex, a drawback that may cause undesirable solutions. In this work, we have attempted to convexify the functional by replacing its non-convex portion by the Charbonnier potential, which is strictly convex and has been reported by authors that it generates stable solutions. Extensive range of experiments have been conducted to demonstrate that our method produces compelling results.