Blind image deblurring via content adaptive method

Abstract

Blind image deblurring aims to obtain a clear image and blur kernel from a blurred image. Most existing methods estimate the blur kernel through the entire image. However, different image information, such as image structure information, smooth area information and noise information, contribute differently to blur kernel estimation. The uniform processing of various image information will reduce the accuracy of blur kernel estimation. In this paper, we propose a new blind deblurring method based on the content-weighted data fidelity term, which can focus more on the sharp edge to restore image structure. Moreover, we construct a new image prior to constrain the weight matrix. However, the content-weighted data fidelity term is a non-convex function. In this work, we introduce the variable splitting method to replace content-weighted matrix, which can be optimized by alternating iteration method. A large number of experiments show that the proposed deblurring algorithm can obtain the best performance on natural images and text images.