Optical Flow Computation Based on Bayesian Selecting Scheme  

作　　者：LU Zongqing XIE Weixin PEI Jihong 

机构地区：[1].School of Electronic Engineering, Xidian University, Xi＇an 710071, China [2]Key Laboratory of Intelligent Information Processing, Shenzhen University, Shenghen 518060, China

出　　处：《Chinese Journal of Electronics》2007年第1期107-110,共4页电子学报（英文版）

摘　　要：Dense optical flow schemes are challenged by the presence of motion boundaries. In state of the art optical flow approaches, over-smoothing of motion boundaries accounts for most of the error. A new optical flow estimation is proposed in this paper, which addresses some issues including the credible estimation of optical flow and the preservation of flow boundaries. Our main contribution is that a Bayesian selecting scheme is adopted to refine the optical flow by a local estimate scheme, which can help to preserve motion boundaries efficiently. Qualitative and quantitative results indicate that the method can produce a credible result.

关 键 词：Optical flow BAYESIAN Cluster. 

分 类 号：TN911.74[电子电信—通信与信息系统] TP242.62[电子电信—信息与通信工程]