提出了一种有效的时空融合视频分割系统。首先利用基于梯度的分水岭算法对图像进行分割,并利用基于时空联合的区域合并方法克服过分割现象,然后结合改进的帧差掩模对目标进行提取。试验结果表明,由于该系统充分利用了图像的颜色信息,能较完整地分割出对象,并且对象边缘更加准确。
An effective video object segmentation algorithm based on spatial-temporal fusion is proposed in this paper. First of all, the image is segmented by watershed algorithm based on gradient image, and over-segmentation is overcome by region merging based on spatial-temporal fusion. Then the target is extracted by frame difference mask, The simulation results show that the proposed system can completely segment video object and its edge is more accurate.