Video Restoration for Shift Varying Motions Stanley H. Chan, and Truong Nguyen

Abstract

There are two types of motion blurs. The first one is caused by stationary scene and moving camera. This is the typical situation of hand shaking and the motion is global. The second one is caused by stationary camera and moving objects. This is the typical situation of capturing a video and the motion is local. Therefore, we investigate methods to restore a blurred video by using motion information. Specifically, using the classical L1 regularized least square framework, we let the point spread function (PSF) being shift variant. Thus the PSF is no longer a block toeplitz with toeplitz block matrix. We investigate efficient preconditioners for the solving this problem. We also propose using adaptive weights to regularize solution around edges and textures.

Results

1. Under Construction ...