Low-pass filtering aiming at noise generated in a contrast enhancement

In this paper, we propose a low-pass filtering process aiming at removing noise and artifacts generated by histogram equalization, while preserving the image signal variations. The filtering is made to provide different levels of smoothing strength by means of cascading stages of simple low-pass filters. A weak smoothing given by the first stage is applied to all the pixels, including those in edge regions, and the pixels located in the flattest regions are processed successively by all the filtering stages to get the strongest smoothing. A binary mask is used in each stage, except the first one, in order to shield pixels in non-homogeneous regions from over-smoothing. Simple algorithms are developed to generate the masks from the input image. The results of the simulation demonstrated that the proposed filtering leads to a good quality of the contrast enhancement in varieties of images and requires a low computation complexity.