Extended Probabilistic Pseudo-Morphology for Real-World Image Denoising

Abstract

Image denoising is an actively researched topic and a multitude of methods have been proposed for this task, including techniques based on mathematical morphology, which is a popular non-linear processing framework developed for binary and grayscale images, based on imposing a lattice structure on the image data. There is no universally accepted extension to the color and multivariate domain and multiple approaches have been developed. Pseudo-morphological operators do not respect all the theoretical properties of classical morphology, but can be successfully used in a variety of applications. In this paper, we present an extension to the Probabilistic Pseudo-Morphology framework by including third-order statistics in the definition of the pseudo-extrema. The approach shows improved performance in the context of a real-world image denoising application in comparison with other color morphological frameworks.