Investigation of performance trade off in motion estimation algorithms on sub-pixel displacement

This paper presents a performance evaluation of two different sub-pixel motion estimation algorithms, one base on Block-Matching and the other based on optical flow to obtain sub-pixel displacement. On block-matching, we focus on block-based full search (FS), three step search (TSS) [1], two dimensional logarithm search (TDL) [2], cross search algorithm (CSA) [3], a new three step search algorithm (NTSS) [4], a novel four-step search algorithm (NFSS) [5], a block-based gradient descent search algorithm (BBGDS) [6], a new diamond search algorithm (DS) [7] and hexagon search algorithm (HS) [8]. And on optical flow, we focus on 2D optical flow block-based full search algorithm (OF), bidirectional confidence based optical flow algorithm (BCOF) [9]. These are evaluated based on mean average error (MAE), peak signal to noise ratio (PSNR), and computational time. These experiment results are comprehensively tested on several standard sequences such as AKIYO, COASTGUARD, CONTAINER, and FOREMAN.