An Efficient Prediction Search Pattern for Half Pixel Motion Estimation

Abstract

Motion Estimation (ME) consists of two steps, the integer pixel motion estimation and the half pixel motion estimation. Many fast integer pixel motion estimation algorithms have been developed to reduce the complexity. However, the half pixel motion estimation requirs huge computational complexity. In this paper, We propose an efficient prediction search pattern for half pixel ME. The proposed search pattern reduces the computational overhead by limiting the number of interpolations of the candidate half pixel points based on the cross center-biased distribution property of Motion Vectors (MV) and the correlations between integer pixel MVs and half pixel MVs. Experiments show that the speedup improvement of the proposed algorithm over a full half pixel search (FHPS), horizontal and vertical directions as references (HVDR), chen’s half pixel search algorithm (CHPS-1) and a parabolic prediction-based fast half-pixel search (PPHPS) can be up to 1.2 ~ 3.3 times on average and the image quality improvement can be better up to 0.01(dB)~ 0.25(dB).