Efficient Impulsive Noise Suppression based on Statistical Confidence Limits

Abstract

Probably, the most well-known vector filter is the vector median filter (VMF) which is based on the theory of robust statistics and performs good noise suppression in color images. However, the VMF is designed to perform a fixed amount of smoothing. This may lead to too much unnecessary substitutions in the input image and, as a result, blurring and loss of image details. In order to avoid this drawback when dealing with impulsive noise, the switching schemes aim at selecting a set of pixels of the input image to be filtered leaving the rest of the pixels unchanged. In this paper, two switching filters which base the selection of the noisy pixels to be filtered on statistical tests are proposed. The proposed filters present good noise suppression while preserving fine image details appropriately. Comparisons to classical and recently introduced impulsive noise multichannel filters are provided. Moreover, the noisy pixel selection techniques are computationally simple, and the filters significantly reduce the computational complexity of the VMF.