基于时空图的轮廓线模型研究

主动轮廓线模型(Snakes)具有能够结合先验知识和图像特征的理论价值,但由于能量局部极值问题的困扰,而鲜已用于实际问题。前人通过改进优化算法来提高其分割效果。文章则尝试从Snakes的外部力入手,引入时空图的光流特征,通过改造能量函数提高分割效果。实验对比验证在复杂背景下,基于时空图的轮廓线模型依然可以有效跟踪运动物体,此时普通Snakes模型必须做大量手动调整。而它同时也解决了帧差等视频分割方法难于识别、提取目标轮廓的问题。时空图轮廓线模型抓住了运动目标的图像特征,在简单的优化算法下依然可以得到较好的分割效果,可以用实时系统。

A Spatio-Temporal Snake Model for Motion Segmentation

Active Motion Models(Snakes)have a great value because of the combination of the image feature and prior knowledge.But ,they are hardly to use because of the problem to optimize the energy function.The people did a lot of work on the optimizing algorithm to improve the models.The paper attempts adding a new outer force for the Snake model to avoid this problem.This new outer force is the angle of the grads of the spatio-temporal picture.By the exper-imentations,we found this new model could well track mobile object on the complex background.In the same condition,other Snake models have invalidated.On the other hand,the problem is also dissolved to recognize the edges by tradi-tional video segmentation algorithms.Spatio-temporal Snake model could track the mobile object rapidly because its opti-mizing algorithm is very simply.So,it could be used in real time system.