Soft Morphological Operators Based on Nonlinear L _P Mean Operators

Abstract

The classical grayscale morphological operators erosion and dilation are characterized by strong nonlinearities, a feature that is not desirable in certain image filtering applications. This paper investigates the use of a subclass of nonlinear mean filters, instead of the classical morphological transformation. The nonlinearities of these filters can be hardened or softened progressively, by means of a controlling parameter. The statistical properties of these “soft” morphological filters are investigated. It will be shown that the nonlinear mean filters behave better than or equally well to the grayscale morphological transformations, for certain types of noise.