Impulse Noise Removal: A comparative Study

The project is aimed at comparing performance of filters for pulse noise removal in images. 22 filters ([1-18]) were implemented on a Matlab platform and tested in terms of: False alarms errors Missing errors Estimation errors due to false alarms and missing Visual quality The comparison results are presented in form of tables in which the filters are sorted in descending order of the quality. Filters The tested filters are divided into classes: Order statistics based filters A histogram based adaptive weighted-median filter [2] A multiple-median related filter. [3] Generalized trimmed mean filter [4] Image block estimation using a block around the median [5] Rank order filter for smoothing and sharpening with automatic parameter selection [9] A recursive minimum-maximum filter [10] Intensity spread based impulse detector for adaptive median [11] Similarity based recursive impulse detector and adaptive weighted-median filter [12] Adaptive α-trimmed and Lpq filters for smoothing an sharpening [13] An adaptive median filter with impulse detector based on gaussian kernel smoothing [14] An adaptive rank-conditioned median (RCM) filter [15] An optimal nonlinear extension of linear filters based on distributed arithmetic [16] Minimum-maximum exclusive mean filter to remove impulse noise from highly corrupted images [18] Modified Intensity spread based impulse detector for adaptive median filter [19] Robust moving average estimation for impulse noise filtering [21] Robust KNN filter for impulsive noise suppression [22] Morphological filters Generalized morphological filters [17] Nonlinear image restoration filters Rank order mean filter and fuzzy impulse noise classifier [1] A fuzzy filter for image corrupted by impulse noise [6] Adaptive impulse noise removal using 2D polynomial approximation and Median based impulse detector [7] Restoration of impulse noise corrupted images using long range correlation [8] Fuzzy filter algorithm using a relaxation algorithm [20]