Bidirectional symmetry and median filter with dynamic smoothness weight on Horn-Schunck optical flow algorithm

For motion estimation, optical flow is an algorithm that uses to determine motion vector (MV) in a level of pixel based over video sequences. Under noisy situations, they corrupt the achievement in reliability of MV on optical flow. To lift the reliability of MV, this paper presents bidirectional symmetry and median filter with dynamic smoothness weight on Horn-Schunck optical flow algorithm. From the experiment, it shows the effective result in our proposed algorithm when various values of smoothness weight are comprehensively considered. We also analyze on how importance of the appropriated value of smoothness weight (α) that should be selected for the best achievement. Several differences in characteristic of conventional sequences are used in our experiment. We also simulate theses sequences with several levels of Additive White Gaussian Noise (AWGN) for performance investigation under noisy environment where Peak Signal to Noise Ratio (PSNR) is a performance indicator.