Comparative Analysis of Noise Filtering Performance of Quadratic Image Filters

Abstract

Quadratic image filters are filters belonging to a subclass of nonlinear model known as Volterra filters. Because of the nonlinear characteristics of images, nonlinear image filters generally produce better results than linear filters. In the present study, performance of the Quadratic image filters for Gaussian noise is examined by comparing with Gaussian filter and Median filter.  For this purpose, the mask weights used were determined by using Differential Evolution algorithm on synthetic training images. Noise added colour test images were filtered using Quadratic image filter using the calculated weights and the results were compared with Gaussian filter and Median image filter.