Experimental study of bidirectional confidential with median filter on global based optical flow model under non Gaussian noise contamination

In this paper, we present a performance study of bidirectional confidential with median filter on global based (Horn and Schunk) optical flow under non Gaussian noise where several robust models on global based optical flow algorithms are used in comparison including Barron kernel, bidirectional confidence based, and median filter for robust motion estimation. The experiment results are simulated over several standard sequences that have different in movement characteristics and speeds with non Gaussian noise in contamination. There are Poisson Noise, Salt & Pepper Noise, and Speckle Noise. The experiment concentrates on Peak Signal to Noise Ratio as a performance indicator where the reconstructed image from the result of motion vector of each method is used to compare with the ground truth in the experimental performance analysis.