作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程

• 图形图像处理 • 上一篇    下一篇

改进的非局部均值滤波算法

郭贝贝,易三莉,贺建峰,苗莹,邵党国   

  1. (昆明理工大学 信息工程与自动化学院,昆明 650500)
  • 收稿日期:2015-06-23 出版日期:2016-07-15 发布日期:2016-07-15
  • 作者简介:郭贝贝(1988-),女,硕士研究生,主研方向为数字图像处理;易三莉(通讯作者),讲师、博士;贺建峰,教授、博士;苗莹,硕士研究生;邵党国,博士。
  • 基金资助:
    国家自然科学基金资助项目“基于PET/CT探测器灵敏度特性的呼吸运动伪影校正研究”(11265007);云南省科技厅省级人才培养基金资助项目“磁共振大脑神经纤维成像技术研究”(KKSY201203030)。

Improved Non-local Means Filtering Algorithm

GUO Beibei,YI Sanli,HE Jianfeng,MIAO Ying,SHAO Dangguo   

  1. (Institute of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650500,China)
  • Received:2015-06-23 Online:2016-07-15 Published:2016-07-15

摘要: 非局部均值滤波算法是一种能够较好保持图像纹理细节的降噪算法,但该算法不能自适应调节滤波参数且滤波结果易产生伪影。针对这些不足,提出一种自适应的且与各向同性算法相结合的非局部均值滤波算法。通过改进的拉普拉斯算子对图像的噪声方差进行估计,并将估计的方差应用于滤波系数的计算中,实现滤波系数的自适应选择。以局部方差作为调整因子,将非局部均值滤波算法与各向同性局部滤波算法相结合,减少滤波图像产生的伪影,提高图像质量。实验结果表明,与非局部均值滤波算法相比,该算法在保留图像纹理细节的同时能够更好地减少伪影,提高图像的峰值信噪比,具有更好的降噪效果。

关键词: 非局部均值滤波, 拉普拉斯算子, 噪声方差, 滤波系数, 各向同性滤波

Abstract: The non-local means filtering algorithm is an efficient denoising algorithm that can restore details of image texture well.However,the algorithm cannot adaptive adjust filtering parameters and the filtering result is easy to produce artifacts.Owing to these shortcomings,an adaptive non-local means filtering algorithm combined with isotropic algorithm is proposed.It estimates variance of image noise by using improved Laplacian operator.The estimated variance of the noise can be applied to the calculation of filtering coefficient,which can realize the automatic selection of filtering coefficient.Local variance is taken as adjustable factor to combine non-local means algorithm and isotropic local filtering algorithm.By this way,the artifact generated by filtering image is reduced and the quality of the image can be improved.Experimental results show that,compared with non-local means filtering algorithm,the proposed algorithm can better reserve details of image texture,at the same time reduce the artifact more efficiently,and improve the peak signal to noise ratio of the image,with a better noise reduction effect.

Key words: non-local means filtering, Laplace operator, noise variance, filtering coefficient, isotropic filtering

中图分类号: