Noise filtering in digital images and approximation theory

摘要

In this paper a kind of equivalence between noise filtering and approximation theory is established. This is done for several noise filtering algorithms for both random and impulse noises. As a result one gets a better understanding of the noise filtering operations. Other enhancing operations may be explained also in this context. We then make use of a special property of the L1 approximation for discrete data to present a simple algorithm for removing impulse noises. Experimental results and comments are given.