Motion deblurring based on hyper-Laplacian model
Taoshen Li · Jisuanji yingyong yanjiu · 2011
This paper first analyzed the heavy-tailed distribution of gradients in natural scenes and proposed an algorithm about motion deblurring based on hyper-Laplacian moldel.It adopted an alternating minimization scheme to optimize the energy equation.Used a look-up table(LUT) for rapidly solving the image deconvolution.Our method can yield good result with medium size pictures in seconds.Compared with the Wiener and Richardson-Lucy methods,our method yields the best overall performance both in the restoration effect and suppression of ringing.