Sorted Min-Max-Mean Filter for Removal of High Density Impulse Noise

Abstract

This paper presents an improved Sorted-Min-Max-Mean Filter (SM3F) algorithm for detection and removal of impulse noise from highly corrupted image. This method uses a single algorithm for detection and removal of impulse noise. Identification of the corrupted pixels is performed by local extrema intensity in grayscale range and these corrupted pixels are removed from the image by applying SM3F operation. The uncorrupted pixels retain its value while corrupted pixel’s value will be changed by the mean value of noise-free pixels present within the selected window. Different images have been used to test the proposed method and it has been found better outcomes in terms of both quantitative measures and visual perception. For quantitative study of algorithm performance, Mean Square Error (MSE), Peak-Signal-to-Noise Ratio (PSNR) and image enhancement factor (IEF) have been used. Experimental observations show that the presented technique effectively removes high density impulse noise and also keeps the originality of pixel’s value. The performance of proposed filter is tested by varying noise density from 10% to 90% and it is observed that for impulse noise having 90% noise density, the maximum PSNR value of 30.03 dB has been achieved indicating better performance of the SM3F algorithm even at 90% noise level. The proposed filter is simple and can be used for grayscale as well as color images for image restoration.