论文部分内容阅读
为了从视频序列中获取初始视频对象,提出了一种改进的基于运动连通性的初始视频对象提取算法.视频中的运动对象高度连通结构化,这就使得运动连通性是适用于视频对象分割的高级特征.据此首先对反映对象的一致性运动的累计帧差图进行尖锐噪声滤除,然后应用自适应阈值算法提取对象运动区域,接着根据运动连通性标记出最大连通区域,通过后处理得到视频对象的分割模版从而有效提取出初始视频对象.对比实验结果表明,该算法能自动、快速、准确地提取出初始视频对象,获得了理想的主客观分割效果.
In order to obtain the initial video object from the video sequence, an improved algorithm for video object extraction based on the motion connectivity is proposed.Motion objects in video are highly connected and structured, which makes the motion connectivity suitable for video object segmentation According to this, firstly, sharp noise is filtered out on the cumulative frame difference map which reflects the consistent motion of the object, and then the adaptive threshold algorithm is used to extract the object motion area, then the maximum connected area is marked according to the motion connectivity, and the post-processing Which can effectively extract the initial video object.Comparison results show that this algorithm can automatically, quickly and accurately extract the initial video object, and obtain the ideal subjective and objective segmentation results.