Image Deblurring Based on Normalized-weighted Total Variation

Abstract

Image deblurring is an important and challenging problem in imaging processing. It aims to restore clear images from degenerated ones caused by camera shake or target motion. The total variation (TV) regularization has been widely used in image deblurring, which needs to be carefully modified because of over-smoothing issue and solution bias caused by the homogeneous penalization. In this paper, a non-blind image deblurring method based on normalized-weighted TV (NWTV) is proposed. The method utilizes a weight vector to indicate the importance of image gradients, and the weight vector is normalized in a range between 0 and 1. The NWTV method can perform well in images with sparse or dense gradients. The performance of NWTV has been compared with several state-of-the-art image deblurring methods including MPTV, TV-ADMM, LO-IG and INSR. Experimental results demonstrate that the proposed method achieves comparative or better performance.