基于运动平滑性的动态图象序列的特征点对应

Feature Point Correspondence in Dynamic Image Sequence Based on the Notion of Motion Smoothness

建立帧间的运动特征点对应是图象序列分析中的一项非常重要且非常困难的工作。根据图象序列的特点，该文提出了一种运动平滑的代价方程，并将运动平滑性与帧间特征点的欧氏距离相结合，以完成图象特征点运动轨迹的确立。由于此文的方法没有建立在运动的刚性约束之上，所以该方法同样适用于各种非刚体运动的特征点对应。

Establishing the motion correspondence of feature points between every two consecutive images is very important and difficult in image sequence analysis. According to the characteristic of image sequences,this paper proposes a cost equation that describes the smooth degree of the feature points. The cost equation and the distance between two feature points in consecutive images are treated as principle in the establishment of motion trajectories. Because this method is not based on the constraint of rigidity, it can solve the correspondence problem of nonrigid motion.