Blind image motion deblurring with L0L0-regularized priors

• A natural image motion de-blurring model is proposed based on an L0L0-regularized prior.
• The proposed L0L0-regularized prior is sparse for natural image gradients and blur kernel.
• An alternating minimization method is used for estimation of latent image and kernel.
• Experimental results demonstrate the approach is effective and efficient.

Blind motion deblurring from a single image has always been a challenging problem. This paper proposes a blind image motion deblurring method which adopts L0L0-regularized priors both in kernel and latent image estimation. A sparse and noiseless kernel and reliable intermediate latent images are generated with this prior constraint. An alternating minimization method is adopted to ensure that latent image and kernel estimation converge at an acceptable time. The proposed method is easy to implement since it does not require any complex filtering strategies to select salient edges which are critical to the explicit salient edges selection methods. The experimental results demonstrate that the proposed method is superior because of the better performance when compared with other state-of-the-art methods and the encouraging results obtained on some challenging examples.