Experimental performance analysis of High Confidence Reliability based on differential optical flow algorithms over AWGN sequences with sub-pixel displacement

Abstract

This paper presents a performance analysis of 3 popular optical flow algorithms (2D optical flow block-based full search algorithm (BOF), Horn-Schunk algorithm (HS) and Lucas-Kanade algorithm (LK)) under the noise conditions. And the confidence based optical flow algorithm for high reliability (CBOF) is applied on these 3 algorithms over different characteristic of standard sequences with several dB of Additive White Gaussian Noise (AWGN). For algorithm of HS and LK, we also applied the kernel model of Barron, Fleet, and Beauchemin (BFB) on these algorithms in our experiment. Especially in HS algorithm, we also investigate the performance on the best average smoothness weight (α) which is prior evaluated by Darun K. and Vorapoj P.. These experiment results are comprehensively tested on several standard sequences such as AKIYO, COASTGUARD, CONTAINER, and FOREMAN that have different foreground and background movement characteristic in a level of 0.5 sub-pixel displacement. Each standard sequence has 4 sets of sequence included an original (no noise), AWGN 25 dB (low noise), AWGN 20 dB, and AWGN 15 dB (high noise) respectively which concentrated on Peak Signal to Noise Ratio (PSNR) as the performance indicator in our experiment.