复杂场景中动目标图象分割的一种快速算法  

Fast image segmentation algorithm for moving objects in complex backgrounds

作　　者：李晓春[1,2] 孙培懋[1,2] 金国藩[1,2] 刘上乾 

机构地区：[1]清华大学精密仪器与机械学系 [2]西安电子科技大学技术物理系

出　　处：《清华大学学报（自然科学版）》1997年第8期19-21,共3页Journal of Tsinghua University(Science and Technology)

基　　金：国防预研基金

摘　　要：提出了一种利用序列图象从复杂背景中分割动目标图象的新算法。首先利用马尔科夫随机场（ＭＲＦ）模型和连续线过程建立较精确的目标函数，采用并行计算方法计算出速度场，然后以此为依据实现对目标图象的并行分割。实验结果表明：该算法不仅能适应多种复杂场景，而且分割精度高、速度快。A new algorithm presented here could be used to segment out moving objects in complex backgrounds on the basis of image sequence analysis. Using a Markov random field(MRF) model with a continuous line process and a parallel computation algorithm, a good estimated motion vector field is achieved. An on line image segmentation method is then described in detail. It is illustrated that the segmentation method is not only suitable to several complex backgrounds but also high in segmentation precision and fast in computation speed.

关 键 词：马尔科夫随机场 运动估计 图像分割 动目标 

分 类 号：TN919.8[电子电信—通信与信息系统]