Improving the denoising performance of Perona Malik filter using adaptive edge indicator

The Perona Malik filter forms the base for various classical diffusion filters. It takes its origin from the heat equation. The images undergo an iterative diffusion process and the noise is removed gradually after each iteration. The basic concept behind various prominent image processing methods like image smoothing, enhancement, denoising etc is mostly based upon PM equation. But the main drawback of the PM diffusion model is the new features appearing in the processed image. In this work the main focus is on alternating between linear and nonlinear smoothing according to the image features. Numerical experiments carried out prove the efficiency of this scheme for edge detection. It also gives better perceptual quality for the processed images in contrast to basic PM or TV model.