Modified Directional and Fuzzy Based Median Filter for Salt-and-Pepper Noise Reduction in Color Image

Abstract

Impulse noise typically emerges due to bit error in image processing and communication progression. The median filtering approach is quite popular for eliminating impulse noise. In recent decades, several updated median filters have been proposed to enhance noise containment and information conservation approaches that have their own shortcomings when detecting and restoring noise. In this paper, we propose a modified directional and fuzzy based median method for filtering color images that are SPN corrupted using noise detection, noise filtering and restoration of noise free image. Noise detection is carried out to classify the pixels as noise pixels or noise-free pixels based on the intensity values. The detected “noise pixels” are subjected to noise filtering in which they are updated to fuzzy based median values of the good pixels number in current taken filtering window depending upon the minimum or maximum directional differences sorted values in all the twelve directions and are restored. Impact of simulations on test images taken demonstrate that the filter proposed beats state-of-the-art impulse noise removal filters namely “Multiclass Support Vector Machine-based Adaptive Filter (MSVMAF)”, “Noise Adaptive Fuzzy Switching Median Filter (NAFSM)”, “Different Applied Median Filter (DAMF)” and “Basic Pixel Density Filter (BPDF)” with regard to “peak signal-to-noise ratio (PSNR)”, “structural similarity index measure (SSIM)” and “normalized mean square error (NMSE)” quality metrics.