Novel edge preserving multiscale filtering method based on mathematical morphology

Abstract

During the course of conventional multiscale morphological filtering, when the noise is filtered, the signals which are smaller than the structuring elements (SE) may be also removed. In this paper, a novel edge preserving multiscale filtering (EPMF) method based on mathematical morphology is proposed. The EPMF method adds the multiscale top-hat transformation and bottom-hat transformation to the conventional multiscale morphological opening and closing filtering. The two added transformations are used to extract and smooth the features which are smaller than the current scale. It is also found that the smaller features have greater possibilities to contain noise particles. Accordingly, the coefficients of top-hat transformation and bottom-hat transformation are modified. Simulation results on the standard gray-level images show that the proposed EPMF method can effectively remove noise and completely preserve the edge of images. It demonstrates better performance than the conventional filtering methods.