基于残差平方和的光流场正则化方法  

Regularization for Optical Flow Field with Residual Sum of Squares

作　　者：范勇[1] 游志胜[1] 张建州[1] 郑文琛[2] 冯子亮[1] 王荣[1] 

机构地区：[1]四川大学图形图像研究所,四川成都610064 [2]四川大学材料科学系,四川成都610064

出　　处：《光电工程》2001年第6期59-62,共4页Opto-Electronic Engineering

基　　金：国家自然科学基金重点项目(69732010)资助

摘　　要：光流场的不连续性是光流场准确估计中最为重要的研究内容。提出利用最小二乘法和加权最小二乘法算法中的残差平方和来对光流场进行正则化,以减少计算窗口中含有多种模式运动的概率,较好地解决了光流场的不连续性,从而使该算法具有鲁棒性。计算结果表明,该算法具有简单实用和运算速度快的特点。Discontinuities in optical flow field is one of the most important research contents in accurately estimation of optical flow field. The regularization for the optical flow field can be carried out by use of residual sum of squares in least-square or weighted-least-square algorithm, in this way the probability of the multi-mode motion containing in the calculation window can be reduced and the discontinuities of optical flow field can be solved, thus the algorithm has robustness. The calculation results show that the algorithm has the characteristics of simplicity, practice and fast operating speed.

关 键 词：光流场 残差平方和 最小二乘法 计算机视觉 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]