PSO based edge keeping suppression of impulses in digital imagery

Abstract

This paper proposes an efficient switching median filter to restore digital images corrupted by high density of random valued impulse noises. The noise detection is performed using all neighbor directional weighted pixels in the 5 x 5 window. Arithmetic absolute differences and intensity of the center pixel is compared with other pixels in the test window to define a noisy pixel. To restore the noisy pixel variable window based median filtering has been done. Particle swarm optimization(PSO), a recent stochastic global optimization technique has been adopted to obtain best fitted parameters of the proposed detection and filtering operators. Simulation results, conducted on a variety of gray scale images clearly exhibit that the proposed operator obtains better results compared to existing directional weighted filters. © 2012 Springer-Verlag GmbH Berlin Heidelberg.