Signal Exclusive Adaptive Average Filter for Impulse Noise Suppression

This paper introduces a novel signal exclusive adaptive average (SEAA) filter that offers good image denoising performance in applications characterized by impulsive or impulse-like noise. The proposed algorithm works well in suppressing impulse noise with noise ratios from 3% up to 60%. We begin by introducing a digital differentiation preprocessing step to quantify the increments in each local neighborhood of the noisy image. A homogeneity level map is then derived by adaptive thresholding and used to designate pixels as noise candidates. The initial selection is refined using a novel connected components labeling algorithm. Finally, the noise is attenuated by estimating the values of the noisy pixels with a linear filter applied exclusively to those neighborhood pixels not labeled as noise candidates. This approach bears similarity to several nonlinear techniques including alpha-trimmed means, selective averaging, and WMMR filters. Simulation results indicate that SEAA is better able to preserve 2-D edge structures from the original image and delivers better performance with less computational overhead as compared to competing nonlinear denoising algorithms