基于分层MRF模型的抗抖动视频分割算法

提出了一种新的抗摄像机频繁抖动的视频分割算法.采用分层马尔可夫随机场（MRF）模型对视频各帧图像进行多分辨率建模，利用视频序列中帧图像的空间关系来提高分割的准确性，通过Gibbs采样算法求得最大后验概率（MAP），从而实现在摄像机抖动情况下对视频目标的准确分割.在强光、多目标以及复杂背景等情况下对视频序列的车辆目标进行分割.经过实验对比，新算法的分割效果明显优于背景累积相减分割算法以及高斯混合模型方法.

Vibration-resisted video segmentation algorithm based on hierarchical Markov random field model

A new video segmentation algorithm for data captured with possible camera vibrations was proposed.The algorithm hierarchically models the frames of video sequences via hierarchical Markov fandom fields(MRF),and achieves accurate segmentation of objects under camera vibrations by using spatial neighboring relationships in each frame.Gibbs sampling was used to obtain maximum A posteriori(MAP).Experiments on video sequence data in the environments of strong illuminations,multiple objects and complex backgrounds show that new algorithm is more efficient than the existing algorithms based on backgrounds accumulated from frame difference and based on Gaussian mixture models(GMM).