14-3 An Impulse Noise Removal Filter based on Gaussian Mixture Distribution Model for Image

Abstract

In this paper, a progressive switching adaptive Wiener Filter (PS/AWF) method based on the Gaussian mixture distribution model(GMM) for images is proposed as a restoration filter for the degraded images by impulse noise. In designing Wiener filter(WF), estimation of covariance of original image becomes key problem. In this method, an universal adaptive WF is designed, based on universal GMM determined with varies training images. Simulation results show that proposed method improves restoration performance compared with conventional adaptive filers.