摘要： 针对运动模糊问题,借助正则化思想,提出基于L0正则化约束以及自然图像梯度分布的先验模型,给出求取模糊核的复原算法。采用T-smooth技术对图像的梯度进行筛选,提取出有利于模糊信息求解的有效边缘,并使用得到的中间结果修正模糊核,从而求得准确的模糊核和清晰图像,利用双边滤波器抑制图像非盲去卷积过程中引入的振铃效应。实验结果表明,该算法具有较好的鲁棒性,可有效地去除运动模糊和抑制振铃效应,得到高质量的复原结果。

关键词: L0正则化, 图像复原, 盲去卷积, 运动模糊, 振铃效应

Abstract: Aiming at the motion blur,a new blind deblurring algorithm is proposed,which is based on the L0 regularization restraints and the prior knowledge of natural image gradient distribution to obtain the real motion kernel.In the proposed methods,T-smooth technology is employed to screen the effective edge which is advantageous to estimate motion information.The refined motion kernel and a clean image are approximated by iterations.The bilateral filter is applied to non-blind deconvolution in order to inhibit the ringing effect.Experimental results demonstrate that compared with the previous approaches,the algorithm can effectively remove the motion blur and suppress the ringing effect.It is also shown that the proposed algorithm generates higher quality deblurring results than the existing algorithms.

Key words: L0 regularization, image restoration, blind deconvolution, motion blur, ringing effect

中图分类号: 

• TP301.6