西安电子科技大学学报 ›› 2016, Vol. 43 ›› Issue (3): 78-84.doi: 10.3969/j.issn.1001-2400.2016.03.014

• 研究论文 • 上一篇    下一篇

高阶SVD和全变差正则的乘性噪声去除模型

霍雷刚1;冯象初1;王旭东2;霍春雷3   

  1. (1. 西安电子科技大学 数学与统计学院，陕西 西安  710071；
    2. 广西师范学院 计算机与信息工程学院，广西 南宁  530023；
    3. 中国科学院自动化研究所模式识别国家重点实验室，北京  100080)
  • 收稿日期:2015-03-19 出版日期:2016-06-20 发布日期:2016-07-16
  • 通讯作者: 霍雷刚
  • 作者简介:霍雷刚(1986-)，男，西安电子科技大学博士研究生，E-mail：leiganghuo@163.com．
  • 基金资助:

    国家自然科学基金资助项目(61271294,61472303,61362029,61379030)；中央高校基本科研业务费专项资金资助项目(NSIY21)

Higherorder singular value decomposition- and total variation- regularized multiplicative noise removal model

HUO Leigang1;FENG Xiangchu1;WANG Xudong2;HUO Chunlei3   

  1. (1. School of Mathematics and Statistics, Xidian Univ., Xi'an  710071, China;
    2. School of Computer and Information Engineering, Guangxi Teachers Education Univ., Nanning  530023, China;
    3. NLPR, Institute of Automation, Chinese Academy of Sciences, Beijing  100080)
  • Received:2015-03-19 Online:2016-06-20 Published:2016-07-16
  • Contact: HUO Leigang

PDF (PC)

215

赞

被引次数

9

摘要：

光滑性、稀疏性和自相似性先验作为自然图像的重要特性被广泛应用于图像去噪．根据高阶奇异值分解和全变差正则的互补性，提出了一种能够同时利用光滑性、稀疏性和自相似性先验的乘性噪声去除新方法．该方法首先采用高阶奇异值分解方法对对数变换后图像中的相似块组进行去噪；然后结合考虑光滑性先验的全变差约束对结果进行迭代优化．实验结果表明，该方法在有效去除乘性噪声的同时，可以更好地保留图像的边缘和纹理区域的细节信息．

关键词: 高阶奇异值分解, 乘性噪声, 全变差, 非局部滤波, 图像去噪

Abstract:

Smoothness, sparsity and self-similarity are the priors widely used in image denoising due to their importance in representing natural images. Motivated by the collaborative roles of higher order singular value decomposition and total variation regularization, a new approach that can simultaneously capture the above priors is proposed in this paper for removing the multiplicative noises. By taking advantages of local adaptiveness, sparsity and self-similarity realized by higher order singular value decomposition, the proposed approach starts with similar-patch-group-wise adaptive denoising on the log-transformed image, followed by the iterative optimization implemented by the total variation constraint which considers the prior of smoothness. Experiments demonstrate the advantages of the proposed approach in removing multiplicative noise and preserving the details near the edges and in the texture area.

Key words: higher order singular value decomposition, multiplicative noise, total variation, nonlocal filter, image denoising

中图分类号: 

  • O29
　陕ICP备09011600号
版权所有 © 2018 《西安电子科技大学学报》编辑部
地址：西安市太白南路2号349信箱(710071) Tel: (029)88202853 Email:xuebao@mail.xidian.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发
Baidu
map