A semiautomatic segmentation algorithm of multiple nonrigid moving objects in clutter background

Tracking multi-nonrigid moving objects in the clutter background is a difficult problem in video object segmentation. This paper presents an improved semiautomatic algorithm. First, we manually draw multiple smallest rectangles to contain the same number of moving objects respectively, called rectangle of interest. Second, the multiple moving objects are automatically obtained and tracked using Hausdorff distance transformation in the video sequence. Third, the binary edges are closed and the video object planes are extracted through a new method presented in this paper. It can segment multiple moving objects from the clutter background, not only solving the problem of uncovered background and aperture, but also closing the convex or concave contours of the objects and extract the VOPs from the video sequence. Experimental results demonstrate the performance of our proposed algorithm.