In single blind motion deblurring

salient edges have been the key to success of kernel estimation.To this end

a new blind motion deblurring algorithm is proposed based on optimal reweighted L1 norm.Firstly

the weighted L1 based blind kernel estimation model is constructed.Then

for reducing the influence of noise and tiny structures

an image smoothing model is introduced into the optimal estimation of weights.A numerical algorithm based on iterative shrinkage-thresholding is also proposed to solve the blind kernel estimation model.At last

a fast non-blind deconvolution method using Hyper-Laplacian priors is utilized to restore the final image.Experimental results on simulated and real-world data demonstrate the superiority of the proposed method.