Jacquard Image Registration Algorithm Based on Spatio-temporal Optical Flow

摘要

Jacquard image registration is a key problem in the analysis of jacquard image sequence. Traditional optical flow based image registration methods are designed to work with static spatial image pairs and does not maintain temporal coherency of them when applied to video sequences. The aim of this work is to develop appropriate spatio-temporal coherency models of optical flow estimation for jacquard image registration. However, incorporating spatio-temporal regularization into an energy optimization framework is challenging due to the constraint penalty terms. A variational framework, together with a coarse-to-fine refinement, is proposed for the optical flow computation by discretizing the associated Euler-Langrange equations and solving the resulting systems of equations in an iterative way. Experiments with real jacquard image sequences demonstrate the favorable performance of the proposed algorithm when compared with the related works.