Local Regularization Robust Optical Flow Estimation with Feedback Adjustment
反馈调节的局部正则化鲁棒光流估计

A new method named Local Regularization Optical Flow Estimation with the mechanism of feedback adjusting was proposed.Some specialties of the global algorithm and local algorithm were merged by this method.The whole field of image was divided into many sub-patches with public boundary,and the distance was defined to decide which sub-patch each point to be within.The optical flow constraint and the smooth constraint were regularized to construct the object function for optimization within the each sub-patch.The intensity offset in the optical flow constraint,as a control parameter,and the peak signal noise ratio to measure the precision,as a output,were used to construct the feedback loop to optimize the optical flow.The experiment result shows that the method make the optical flow more precise,and it make sense to improve the performance of existent optical flow algorithms.