面向视频的人物前景分割研究

Foreground segmentation from videos

针对目标分割中存在的背景噪声和阴影干扰问题,提出一种改进的分割方法:利用码书背景建模对视频进行分割后,结合梯度因子、Sobel算子对分割结果进一步约束。实验结果表明,该方法可以有效地分割出前景目标,抗干扰能力强。

For the problems of background noise and shadow interference in the object segmentation,this paper proposed an improved target segmentation method.After using the codebook background model segmented the video,it constrained the segmentation results for further by combining the gradient factor and the Sobel operator.Experimental results show that the method can effectively segment the foreground targets and has strong anti-interference ability.