计算机工程

• 图形图像处理 • 上一篇    下一篇

基于改进K-SVD 和非局部正则化的图像去噪

杨爱萍,田玉针,何宇清,董翠翠   

  1. (天津大学电子信息工程学院,天津300072)
  • 收稿日期:2014-04-30 出版日期:2015-05-15 发布日期:2015-05-15
  • 作者简介:杨爱萍(1977 - ),女,副教授、博士,主研方向:图像处理;田玉针,硕士;何宇清,讲师、博士;董翠翠,硕士。
  • 基金项目:
    国家自然科学基金资助项目(61372145)。

Image Denoising Based on Improved K-SVD and Non-local Regularization

YANG Aiping,TIAN Yuzhen,HE Yuqing,DONG Cuicui   

  1. (School of Electronic Information Engineering,Tianjin University,Tianjin 300072,China)
  • Received:2014-04-30 Online:2015-05-15 Published:2015-05-15

摘要: K-奇异值分解(K-SVD)算法在强噪声下的去噪性能较差。为此,提出一种新的图像去噪算法。使用相关 系数匹配准则和噪声原子裁剪方法改进传统K-SVD 算法,提高原算法的去噪性能,将非局部正则项融入图像去噪 模型,并采用非局部自相似性进一步改善图像的去噪效果。实验结果表明,与传统K-SVD 算法相比,该算法在提 高同质区域平滑性的同时,能保留更多的纹理、边缘等细节特征。

关键词: 图像去噪, 稀疏表示, 奇异值分解, 正交匹配追踪算法, 字典优化, 非局部自相似性

Abstract: In view of the poor performance of the K-Singular Value Decomposition(K-SVD) denoising method,a new algorithm is proposed. The denoising performance is improved by the refined K-SVD method with the help of the correlation coefficient matching criterion and dictionary cutting method. By combining the non-local self-similarity as a constrained regularization into the image denoising model,the performance is further enhanced. Experimental results show that compared with traditional K-SVD method,this algorithm can effectively improve the smoothness of homogeneous regions with preserving more texture and edge details.

Key words: image denoising, sparse representation, Singular Value Decomposition ( SVD ), Orthonomal Matching Pursuit(OMP) algorithm, dictionary optimization, non-local self-similarity

中图分类号: