计算机工程与应用 ›› 2016, Vol. 52 ›› Issue (18): 173-178.

• 图形图像处理 • 上一篇    下一篇

基于l1/2范数正则化的图像重建方法

查志远,刘  辉,尚振宏,李润鑫   

  1. 昆明理工大学 信息工程与自动化学院 计算机系,昆明 650500
  • 出版日期:2016-09-15 发布日期:2016-09-14

Image reconstruction method based on l12 norm regularization

ZHA Zhiyuan, LIU Hui, SHANG Zhenhong, LI Runxin   

  1. Department of Computer, School of Information Engineering and Automation, Kunming University of Science and Technology, Kunming 650500, China
  • Online:2016-09-15 Published:2016-09-14

摘要: 为了利用[l1]范数保持图像边缘信息的优势,并兼顾[l2]范数对图像平坦区域噪声抑制的特性,提出了一种自适应范数混合模型——[l12]范数正则化方法。相比于经典的[l1]范数正则化方法,该方法能够得到更加稀疏的解,同时相比于传统去噪方法,该方法对自然图像的长尾分布噪声具有比较理想的去除效果。还针对范数混合模型中噪声的分布的自适应变化,设计了一种自适应收敛准则迭代方法,该方法可以有效地减少迭代次数。实验结果和分析验证了混合模型在图像重建效果和计算效率方面的有效性。

关键词: 图像重建, 自适应范数混合模型, 正则化, [l12]范数, 自适应收敛准则

Abstract: The purpose of keeping image edge preservation by [l1]norm, while suppressing noise by [l1]norm in the field of smooth, an adaptive norm mix model——[l12]norm regularization method is proposed. Compared with traditional [l1]norm regularization method, this method can get much sparse solution and has an ideal effect on removing the heavy-tail distribution noise. In addition, due to noise distribution will adaptively change, an adaptive convergence rule is proposed, which can effectively decrease the iteration numbers. Experimental results and analysis show that the proposed method can get the ideal restoration result and ?computational efficiency effectively.

Key words: image restoration, adaptive norm mix model, regularization, [l12] norm, adaptive convergence rule