Blind Image Deblurring Based on Local Rank

Conventional algorithms for blind image deblurring are often inaccurate at blur kernel estimation, and the recovery effect is far from perfect. To address this, we propose a single-image blind deblurring method based on local rank. For this, we first impose adaptive threshold segmentation on a conventional local rank transform, which is subsequently used to construct a novel model for blind image deblurring. Next, a half-quadratic splitting method is adopted to estimate the blur kernel and an intermediate latent image, in alternating iterations. Finally, the desired latent image is obtained by linear combination of the hyper-Laplacian model and the total-variation-l₂ model, where the weights are calculated from the adaptive local ranks. Experimental results using public datasets show that the proposed approach can accurately estimate the blur kernel and effectively suppress ringing effects.