基于静止分割和运动矢量的对象跟踪

WatershedObject Tracking Based on Static Segmentation and Motion Vector

运动对象的分割和跟踪是视频图象处理的一个很重要和困难的部分。提出一种非常有效的对象跟踪算法,能够检测到当前帧给定区域的对象并跟踪,不需要事先了解此对象的空间时域的信息也不需要进行镜头运动补偿。静止分割采用算法,而运动矢量采用块-,Watershed匹配法。尽管所要跟踪对象会经历某些形变,依旧能获得较好的结果。

This paper presents an object-tracking algorithm. The proposed tracking algorithm helps to detect the object present in a selected region in the current frame with no previous spatio-temporal information about the object and no camera motion compensation. In here, the Watershed method is used for static segmentation, while the block matching method is used for motion vector. A very good result can be obtained, though the tracked object undergoes some kinds of transformation.