Reduction of Impulsive Interference in Color Image Applications by Using Nonlinear Order Statistics Filters
Volodymyr I. Ponomaryov, Alberto Jorge Rosales-Silva · 2004
Usually the image or video data are degraded through the insufficient resolution ability caused sensor. Random noise also can significantly distort an image or video sequence quality. The paper presents an analysis of different non-linear filtering techniques used to suppress impulsive noise and preserve fine color image details. Directional processing, nonparametric approach and order statistics filters have been investigated in here. We also present a novel filtering scheme that can be able to decrease the noise influence by filtering of corrupted color image or video sequence and provide preservation of fine details. The before presented algorithm MM-KNN and another novel ones RM type filters (WM-KNN, ABSTM-KNN) have been adapted to color imaging and investigated in here. Such a RM-KNN filter can provide color detail preservation and uses the combination of the redescending M-estimators with one from median, Wilcoxon or two other designed estimator for KNN filter calculating the robust point estimate of the pixels within the filtering window and provide impulsive noise rejection.. We have applied in redescending M-estimator different influence functions: the simplest cut, Hampel's three part redescending, Andrew's sine, Tukey biweight, and Bernoulli ones. Several criteria used to compare the quality of image processing are very important to characterize the level of noise suppression obtained in each algorithm, edge, fine details preservation, and color perceptual, and chromaticity errors. Such the criteria were the Pick Signal Noise Relation (PSNR), Mean Absolute Error (MAE), Normalized Color Difference (NCD), and Mean Chromaticity Error (MCRE). The analyzed algorithms have been implemented on DSP Texas Instruments DSP TMS320C6701 for proving the possibility to process an image in real time and to obtain the values of processing time when different intensity noise corrupts a color image or sequence.