An alternative robust optical flow on Dynamic Smoothness Weight (α) of Horn-Schunk Algorithm using median filter with sub-pixel displacement

Traditionally, the smoothness weight (α) of Horn-Schunk optical flow algorithm (HS) is an important parameter of the optical flow and directly impacts to the performance of the HS algorithm. This paper presents an alternative robust optical flow on dynamic smoothness weight (α) of Horn-Schunk algorithm using median filter with sub-pixel displacement (RDHS). We also present an investigation analysis of sub-pixel optical flow on original Horn-Schunk algorithm over various smoothness weights or Dynamic Smoothness Weight of Horn-Schunk Algorithm (DHS) to investigate the value of the smoothness weight that return the best in Peak Signal to Noise Ratio (PSNR) on each frame of video sequence for comparison. We investigate the performance over the Additive White Gaussian Noise (AWGN) at several noise power levels (such as AWGN at 25 dB (low noise), AWGN at 20 dB, and AWGN at 15 dB (high noise) respectively). These experiment results are comprehensively tested on several standard sequences such as AKIYO, COASTGUARD, CONTAINER, and FOREMAN that have difference foreground and background movement characteristic in which PSNR is concentrated as the performance indicator. We also investigate the relationship of the movement characteristic in the sequence and noise level with the best smoothness weight.