基于视差和变化检测的立体视频对象的实时提取

Timely stereo video object segmentation based on disparity and moving detection

随着立体视频的发展,立体视频对象提取技术也成为研究的热点。本文提出一种基于视差和运动变化检测的立体视频对象实时提取算法:结合立体视频的视差、运动和边缘等信息,由立体匹配得到视差图,对视差图初次提取得到二值图像,再运用对称差分法变化检测,根据对象的运动速度实时选取用于变化检测的视频帧,结合运动信息提取出最终的立体视频对象。实验表明该算法能够实时地从复杂背景中提取出视频对象,并可以保证在视频停止变化时保持视频对象的连续输出。

With the rapid development of stereo video,video object segmentation technology has attracted people＇s attention.In this paper,combining depth,motion and edge structure,a depth and spatiotemporal information based algorithm for stereo video extraction is proposed.Disparity is obtained by stereo matching,approximate video objects are obtained from the extraction of disparity map.Moving object extraction performs in the regions derived from the first extraction procedure,based on a twice frame difference algorithm,which can timely select the frames used in the detection based on the speed of object.Then,motion objects and edge features are fused to modify the boundaries of objects.Finaly,the video object is obtained.The results of experiment show that this algorithm can quickly and timely extract video object from the previous frame.The video object output will act as the current video frame video object output to ensure the continuity of the video object segmentation.