A Comprehensive Statistical Scrutiny of HDT Dissimilarity for Localizing and Restoring Impulsive Noisy Images

This scrutinized report presents the comprehensive scrutiny of noisy restoration technique based on HDT (Hard Decision Threshold) dissimilarity, one of the most practical and efficient dissimilarity measurement technique for localizing the impulsive noisy pixels, for RVIN (Random Value Impulsive Noise). The main contribution of this scrutinized report is the determination of the optimal window size and optimal HDT threshold of the noisy restoration technique based on HDT. From many standard tested images, the comprehensive scrutiny illustrates that the obvious difference of the HDT dissimilarity between the genuine pixel and the noisy pixel. Moreover, the statistical relationship between HDT dissimilarity and the cluster size (3x3, 5x5 or 7x7) is simulated for determine the optimized cluster size from performance perspective. Finally, the statistical relationship between the denoising performance and the Hard Decision Threshold (HDT) is simulated by using many tested images for determine the optimized threshold.