Micro order emphasis: a probability cohesion approach to image noise reduction and enhancement

The author proposes a generalized image enhancing scheme which is based on the spatially consistent cohesion of image features in their feature space. One principle of this method is the reduction of additive noise through cohesion of a blurred density function of features. Another principle is the derivation of spatially consistent global image correction from the micro image corrections obtained by the cohesion principle. In order to realize the principle, the author proposes an iterative Bayesian estimation procedure which is recursing between an image space and a feature space using properly designed mapping and remapping methods. It is shown theoretically and experimentally how the two techniques used by the method (i.e. edge line-enhancing noise reduction and texton-enhancing smoothing) work