Asymmetric directional diffusion based image filtering and enhancement

This paper introduces a new method for directional diffusion based image filtering and enhancement. The main novelty of our approach consists in considering asymmetric orientation information for steering a diffusion process along the main directions of the constituent structures of an image. The diffusion process is obtained by generalizing an existing approach defined on a symmetric framework. The increased efficiency of the method for the preservation and enhancement of junctions and corners, coupled with good noise filtering capabilities, is illustrated through an experimental setup involving both synthetic and real images.