A Novel Flatted Hexagon Search Pattern for Fast Block Motion Estimation (TA-P3)

Author(s) : Thou-Ho (Chao-Ho) Chen (Department of Electronic Engineering, National Kaohsiung University of Applied Sciences, Taiwan) Yi-Fan Li (Department of Electronic Engineering, National Kaohsiung University of Applied Sciences, Taiwan)

Abstract : The paper proposes a cost-effective fast black motion estimation based on the flatted hexagon search (FHS). Generally the probability of horizontal motions is significantly more than that of vertical motions in real-world video data captured. This motivates that a novel strategy in which the covering range of a search pattern should be enlarged especially on the horizontal dissection, is introduced in to increase the rate of optimal matching. Based on same searching process as the hexagon-based search (HEXBS) algorithm, the proposed method bears a speed improvement rate (SIR) of ranging from 21% to 25% and ¡V0.7% to ¡V7.7% over the diamond search (DS) algorithm, and the HEXBS algorithm, respectively. In point of probability improvement of finding true motion vector per block, the FHS, HEXBS and DS algorithms have 96.6%, 94.3% and 95.2% in Salesman sequence respectively, and 81.9%, 69.3% and 87.3% in Foreman sequence respectively, when compared to the full search algorithm.

Menu