A window adaptive hybrid vector filter for color image restoration

Abstract

A novel window adaptive hybrid filter for the removal of different types of noise which contaminate color images is proposed. At each pixel location, the image vector is first classified into several different signal activity areas through a modified quadtree decomposition on the luminance of the input image, with only one empirical parameter required to be controlled. Then, an optimal window and weight adaptive vector filtering operation is activated for the best tradeoff between noise suppression and detail preservation. The proposed filter has demonstrated superior performance in suppressing several distinctive types of color image noise, which include Gaussian, impulse, and mixed noise. Significant improvements have been achieved in terms of standard objective measurements as well as the perceived image quality.