Abstract: The segmentation of objects in video sequences constitutes a prerequisite for numerous applications ranging from computer vision tasks to second-generation video coding.A video sequence can be represented as a three-dimensional volume with two spatial and one temporal-coordinates.From this perspective，motion can be estimated by analyzing orientations of local gray-value structures using the three-dimensional structure tensor.In the paper，the correlation between the motion objects and gray value changing is analyzed，the spatial-temporal structure tensor is proposed and used to segment the motion objects.Experimental results demonstrate that the spatial-temporal structure tensor is efficient for the segmentation of motion object in contrast with the traditional structure tensor.

摘要： 视频中运动物体的分割是众多视频应用的前提。视频可以看作是三维时空中的对象，通过邻域的灰度值的结构方向来分析视频的运动。提出了时空结构张量进行运动物体分割的方法。首先分析了视频中物体的运动与灰度的变化的关系，然后删除了三维结构张量中与时间无关的信息，并将此时空结构张量用于视频中运动物体的分割。实验表明时空结构张量运动物体分割方法相对于传统的结构张量，分割准确，计算简单。