Since 1998, the Bilateral filter (BF) is worldwide accepted for its performance in practical point of view under Gaussian noise however the Bilateral filter has a poor performance for impulsive noise. Based on the combining of the Rank-Ordered Absolute Differences (ROAD) detection technique and the Bilateral filter for automatically reducing or persecuting of impulsive and Gaussian noise, this Trilateral filter (TF) has been proposed by Roman Garnett et al. since 2005 but the Trilateral filter efficiency is rest absolutely on spatial, radiometric, ROAD and joint impulsivity variance. Hence, this paper computationally determines the optimized values of the spatial, radiometric, ROAD and joint impulsivity variance of the Trilateral filter (TF) for maximum performance. In the experiment, nine noisy standard images (Girl-Tiffany, Pepper, Baboon, House, Resolution, Lena, Airplain, Mobile and Pentagon) under both five power-level Gaussian noise setting and five density impulsive noise setting, are used for estimating optimized parameters of Trilateral filter and for demonstrating the its overall performance, which is compared with classical noise removal techniques such as median filter, linear smoothing filter and Bilateral filter (BF). From the noise removal results of empirically experiments with the highest PSNR criterion, the trilateral filter with the optimized parameters has the superior performance because the ROAD variance and joint impulsivity variance can be statistically analyzed and estimated for each experimental case.
Decision-Based Marginal Total Variation Diffusion for Impulsive Noise Removal in Color Images
