Anisotropic Diffusion For Smoothing: A Comparative Study

Anisotropic diffusion is a powerful image processing technique, which allows simultaneously to remove noise and to enhance sharp features in two and three dimensional images. Anisotropic diffusion filtering concentrates on preservation of important surface features, such as sharp edges and corners, by applying direction dependent smoothing. This feature is very important in image smoothing, edge detection, image segmentation and image enhancement. For instance, in the image segmentation case, it is necessary to smooth images as accurately as possible in order to use gradient-based segmentation methods. If image edges are seriously polluted by noise, these methods would not be able to detect them, so edge features cannot be retained. The aim of this paper is to present a comparative study of three methods that have been used for smoothing using anisotropic diffusion techniques. These methods have been compared using the root mean square error (RMSE) and the Nash-Sutcliffe error. Numerical results are presented for both artificial data and real data.