A Recursive Gaussian Weighted Filter For Impulse Noise Removal

A recursive filter for effective suppression of impulse noise is presented in this paper. The proposed work estimates the noise corruption level and the position of impulses in the first stage. The appropriate filter parameters for a detail-preserving restoration of the corrupted pixels are determined based on these estimations. The filtering process assumes a Gaussian spatial profile in the neighborhood of corrupted pixel and interpolates accordingly. A normalized, truncated, trimmed and scaled Gaussian weighting function is proposed for the coefficients of interpolation. Extensive simulations show that the proposed filter restores fairly well even the images that are highly corrupted.