基于图像序列的非刚性物体运动形变程度估计方法

摘要

形变程度的定量分析是非刚性物体运动研究中一个重要问题.现有的形变程度估计方法一般都假设特征点位置误差是各向同性,并且不考虑特征点数据丢失的问题.通常事实并非如此,所以本文提出一种同时考虑图像的方向性误差和特征点数据丢失问题的形变程度估计方法.理论研究和人活动图像序列的实验表明,本文提出方法十分有效.%Quantitative characterization of the deformation degree is very important to properly estimate the non-rigid shape motion. The existing deformation degree estimation methods assume the uncertainty in a point position isotropicly and not consider the missing data problem. The assumption are usually not the case, so we provide a method for deformation degree estimation taking the factor of directionality uncertainty and missing data into account. Theoretical analysis and experimental results from human activities image sequence demonstrate the feasibility of the proposed approach.