基于马尔可夫随机场的运动分割研究

运动对象的分割技术一直是图像处理和计算机视觉领域的重要研究课题。采用一种将运动估计方法与马尔可夫随机场(MRF)模型相结合的运动分割方法。采用鲁棒统计技术与误差模型相结合构成运动估计的目标函数,运动模型为仿射运动,通过过松弛算法获得每种运动的运动参数;根据误差最小原则确定运动对应区域的初值,采用马尔可夫随机场(MRF)模型对运动估计结果进行平滑去噪。最后给出了该方法在通用图像实例上的实验结果。

Research on motion segmentation based on robust error model and Markov random field

Segmentation of moving objects is always an important researching subject in image processing and computer vision.This paper uses an algorithm of motion segmentation combined motion estimation with Markov Random Field(MRF).The objective function of motion estimation consists of error model and robust statistical technology.Affine model is chosen as motion model whose parameters are derived by over-slack algorithm.Initial regions are obtained according to minimum error criterion then smoothed by MRF frame to remove noise.The experimental results on common-used image sequences of this method are presented in the paper at last.