Non-blind Motion Deblurring Using L1 Data Fidelity and L0 Sparse Representation
Guodong Wang · Journal of Information and Computational Science · 2014
Motion deblurring is very useful in image processing and attracted much attention in recent years. Nonblind deconvolution is the key component in motion deblurring approaches when the kernel is estimated. Non-blind deconvolution is an ill-posed problem, we introduce L0 sparse prior term for motion deblurring because it is a better smoothing term in edge-preserving smoothing approaches. In this paper, L1 norm based data fidelity term is also introduced for the superior in edge and corner preserving. For the ease of equation solving about L1 norm, the Split Bregman method is introduced. Extensive experiments on image deblurring with different blurs indicate that the proposed approach is stable and valide.