Single-image Motion Deblurring Using Charbonnier Term Regularization

摘要

The blind deconvolution algorithm of motion blur image is one of very hot research in the image processing field currently. In order to get the sharp image and point spread function (PSF), variational method is used. In this paper, we select TVL2 term as data term and propose the Charbonnier term as smooth term. Normalized Charbonnier term can lead the energy decrease while solving the equation and make the energy equation get its convergence much faster. In order to reduce the complexity of the solving the equation, a fast method called Split method is introduced. Not only Charbonnier term has the strong local adaptability which can select large gradient information of the image, exclude small disturbance on the boundary and enhance the selected edges, but also it have a faster convergence speed get the sharp image and point spread function quickly. Experiments demonstrate the validity of the proposed method.