基于标号场的光流法二维运动估计  

TWO-DIMENSION MOTION ESTIMATION BASED ON LABELING FIELD AND OPTICAL FLOW FIELD

作　　者：杨勇[1] 王桥[1] 吴乐南[1] 

机构地区：[1]东南大学无线电系,南京210096

出　　处：《电子与信息学报》2001年第12期1321-1325,共5页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(60072013;60002008)

摘　　要：该文提出了一种改进的二维运动估计光流算法。在光流基本约束和全局平滑性约束的基础上,引入二维运动标号场,对Horn-Schunck算法进行了改进,不仅加快了算法的收敛速度,而且提高了运动估计的准确性,减轻了遮挡边界问题的影响。实验结果证明了该算法的有效性。In this paper, an improved algorithm combining optical flow field with labeling field is presented for estimating the two-dimension motion. Based on the basic constraint and the global smooth constraint, a labeling field is introduced, by which Horn-Schunck algorithm is modified. This method not only fastens the convergence of the algorithm, but also improves the accuracy and weakens the effect of the occluding boundary on the motion estimation. The experimental results show that it is a suitable method.

关 键 词：光流场 标号场 二维运动估计 光流法 图像处理 信号估计 

分 类 号：TN911.73[电子电信—通信与信息系统]