Decomposition and separated adaptive digital processing of degraded images with an visual quality criterion

In this paper, image restoration in the context of visual quality criterion and two-component image model is considered. Taking a signal as a two additive components is a very useful way for modeling the two fundamental features of images : edges and textures. We propose a new algorithm which splits the luminance signal of complex images into plateau luminance signal and texture. The method is based on 2-D median filtering with adaptive support shape and offset as estimator of the local-luminance mean value. Using an objective visual quality criterion deduced from a model of human perception, separate adaptive 2-D linear filtering are performed on the two components according to specific action the degradations have made on.