Abstract: Aiming at the images with neat boundaries and the additive noise satisfies the condition that its distribution is nonGaussian such as uniform, salt & pepper, Laplace and so on, a GNC method for image restoration based on l1-nonconvex nonsmooth function is presented taking the ideas of fitting gradually and variables conversion. According to the original nonconvex nonsmooth function, it establishes a series of fitting functions, which can get better initial value of original problem, meanwhile variable conversion is used effectively to deal with the nondifferen- tiability of l1-nonconvex nonsmooth function. Experimental results show that the method has higher restoration performance.

Key words: l1-nonconvex, nonsmooth, nonGaussian, neat boundary, penalty potential function, image restoration

摘要： 针对有整齐边界且加性噪声满足均匀、椒盐、拉普拉斯等非高斯分布的图像，根据逐步拟合及变量转换思想，提出一种基于l1-非凸、非光滑函数的逐步非凸图像恢复方法。根据原始非凸、非光滑函数建立一组拟合函数，获得原始问题的优良初始值，利用变量转换解决l1-非凸、非光滑函数的非可微性问题。实验结果表明，该方法恢复的图像效果较好，峰值信噪比较高。

关键词: l1-非凸, 非光滑, 非高斯, 整齐边界, 惩罚函数, 图像恢复

CLC Number: 

• TN911.73