用于立体视频会议系统的立体对象分割与跟踪算法

基于对象的立体视频编码压缩技术能在立体视频会议系统中得到很好的应用，从立体视频信号中正确分割出立体视频对象是基于对象的立体视频编码压缩的一个前提条件，基于立体视频会议图像序列的时空特性和左右通道间的视差特性，提出了一种立体视频对象分割与跟踪算法，首先利用空域分割和运动检测相结合的方法，提取左通道中的运动物体；然后，提出一种左右通道间基于边缘轮廓的二级视差匹配算法，并根据已分割的左通道运动对象提取右通道的视频运动对象；最后利用对象边界轮廓的跟踪方法对后续图像中的运动对象进行快速跟踪，实验结果说明文中算法能够应用于立体视频会议图像序列的立体对象分割与跟踪。

An Algorithm of Object Segmentation and Tracking for Stereoscopic Videoconferencing System

Object based stereo video coding is feasible for stereoscopic videoconferencing systems In this paper, a new algorithm for stereo object segmentation and tracking is proposed for stereoscopic videoconferencing based on temporal spatial characteristics between frames of left channel and disparity features between the left and right channels First, the object is extracted from left channel with an improved watershed algorithm and high order statistics Then, a contour correspondence algorithm based on two level disparity estimation is proposed for object extraction from the right channel according to the extracted object in the left channel Finally, an object tracking algorithm is given for object segmentation in the following frames of stereo video by means of contour matching Experimental results show that the proposed algorithm can segment and track video object pair in image sequence of stereo videoconferencing efficiently