考虑镜头突变和视频对象消失情况下分割视频对象(VOP)

Segmentation of Video Object Plane ConsideringShot Break and Object Disappearance

将视频系列分割成语义上有意义的对象即视频对象(VOP:VideoObjectPlane),是实现基于内容压缩编码标准MPEG-4的关键。本文分析T.Meier分割方法的不足:当镜头突变时,形态学波滤和变化检测掩码方法不能检测到独立运动部件;当视频对象消失时,Hausdorff距离匹配实效。本文通过检测镜头突变和判断视频对象的消失,克服了T.Meier种分割方法的不足。检测镜头突变方法简单、计算量不大且有较好的鲁棒性;通过慢变部分是否存在判断视频对象是否消失。

Segmentation of semantic video object plane (VOP) through video sequence is a key to standard MPEG-4 withcontent-based video coding. This paper analyses weakness of segmentation approach in T.Meier's paper,whichapproaches based in Morphological Motion Filter and Change Detection Mask can't detect independent motioncomponent (IMC) when shot breaks, and match of Hausdorff distance is invalid when object disappears. Thispaper has overcome the weakness by detection of shot break and object disappearance. The approach of shotbreak is simple, computationally cheap and robust. Disappearance of object is decided by nonexistence of slowchange component.