A NOVEL DIFFUSION FILTER FOR IMAGE RESTORATION AND ENHANCEMENT

We are proposing a new filter for image restoration and enhancement tasks that relies on image orientation analysis techniques for steering and directing smoothing or enhancement processes. The filter is developed under the Partial Differential Equations (PDE) framework using asymmetric orientation estimation operators and it has better junction preservation and noise removal properties than existing classical PDE based methods. We employ an experimental setup involving several computer generated images and a statistical interpretation of the results for proving the efficiency of the method in processing images composed of directional textures and degraded by additive Gaussian noise. The algorithm is compared also with state of-the-art non-PDE based methods in filtering images containing oriented patterns.