引入光流法的活动轮廓模型

Active Contours Guided By Optical Flow

本文在活动轮廓模型的基础上,引入了“图象统计势能”和光流法,提出了一种新的图象分割方法。该方法改进了活动轮廓模型的一些不足,能准确地检测出物体轮廓边界,且分割结果对初始位置不敏感,同时提高了对运动速度较快的物体轮廓检测的正确性。实验结果证实了该方法的有效性。

A method for segmentation and tracking of moving object in image sequences is presented. The active contour model is a common method for tracking moving objects. But it is sensitive to the initial contour and difficult to track fast moving objects. The concept of Image Statistical Energy and Optical flow are employed to dissolve the problem of tracking fast moving objects. The algorithm was tested on playing badminton sequences. The contour successfully was tracked as boundaries of sportsman.