Motion deblurring from a single image using multi-layer statistics priors

In order to remove camera shake from a single photograph, we present a new deconvolution approach based on multi-layer statistics priors. Our approach is based on the three observations: the estimation of image gradients distribution can affect the deblurring quality; homogeneity region of the blurred image can serve to reduce ringing artifacts; regions with different homogeneities are with different statistics even for a single scene. Based on these observations we propose a multilayer deblurring model and the experimental results show that our model can reduce ringing while improve deblurring quality.