基于分块处理的条件随机场视频分割算法

Video Segmentation Algorithm Based on Partitioning-Processing Conditional Random Field

为降低传统基于条件随机场(CRF)的视频分割的时间复杂度,提出了一种基于视频分块处理的CRF视频分割算法.该算法利用像素空间相关性对原视频帧进行分块处理形成新视频帧,然后使用CRF对新视频帧进行分割,最后根据初始分割、前一帧分割结果和当前帧CRF分割结果实现最终的分割.实验结果表明,在不明显增加误分割率的前提下,文中算法能有效降低时间复杂度.

In order to reduce the time complexity of the traditional video segmentation based on the conditional random field（CRF）,a partitioning-processing algorithm is proposed.In this algorithm,the spatial correlation of pixels is used to partition the original frames into some new ones that are then segmented by the CRF,and the final segmentation is realized according to the initial segmentation,the results of the previous frame and the results of the current frame after CRF segmentation.Experimental results show that the proposed algorithm effectively reduces the time complexity without obvious increase in the false rate of video segmentation.