A new adaptive switching approach for impulse noise removal from color images

Abstract

A new adaptive switching approach is presented in which the detection of noise is based on entropy of the pixels. Then the detected noisy pixels are replaced with the output of a fuzzy weighted filter. With a little increase in computational complexity over the basic Vector median filter and its varieties, this technique works well both in lower and higher noise ratios. Simulation results show that this method outperforms many other existing nonlinear filters in terms of noise reduction and fine details preservation. This technique can also be extended for images corrupted with Gaussian noise and mixed Gaussian and Impulse noise.