运动估值块匹配算法的一种改进

A Modified Block-matching Algorithm for Motion Estimation

分析了在运动估值/运动补偿中的块匹配算法(BMA,Block MatchingAlgorithm)中常用的MAD(MeanAbsoluteDifference)算法的优点及其存在的缺陷,提出了一种改进的结合MAD优点的基于最佳门限的像素匹配归类算法PMCAT+MAD(Pixel MatchingClassificationAlgorithminanOpti malThresholdEstimatedbyMAD),该算法采用MAD方法寻求用于归类的最佳门限,对最佳门限的确定有自适应性.通过理论分析和实验证明,该算法具有接近MSD(MeanSquareDifference)的精度而又保持MAD的计算复杂度小的优点.

<Abstrcat> In this paper,the authors discuss the advantage and shortage of mean absolute difference(MAD) algorithm,which is often used in the block-matching algorithm(BMA) for motion estimation(\)motion compensation,and present a modified pixel-matching classification algorithm in an optimal threshold (PMCAT), which combines the advantages of MAD algorithm.When the algorithm applying MAD method is used to solve the optimal threshold used in classification,it has self-adaptability in determineing optimal threshold.Theory analysis and experiment prove that the algorithm has the merit of approching the accuracy of MSD(mean square difference)and that of relatively easy calculation like MSD.