首页|基于非局部自相似字典学习的图像混合噪声去除算法

基于非局部自相似字典学习的图像混合噪声去除算法

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
  • 维普
高斯噪声和泊松噪声是影响图像质量的主要噪声源,为了去除这些噪声,大量的图像去噪算法被提出,但这些算法往往局限于去除单一的高斯噪声或泊松噪声.由于字典原子能够自适应的表示图像的结构特征,结合高斯-泊松混合噪声的统计特性,提出一种基于字典学习和非局部结构聚类的高斯-泊松混合噪声去除方法,在稀疏表示的框架下学习字典并重构无噪图像.考虑到图像在非局部范围内存在自相似性,通过对非局部的相似结构进行聚类,将此自相似性作为去噪目标函数的正则项,以提高去噪性能.实验结果表明:提出的算法能够有效的去除图像中的高斯-泊松混合噪声.
Mixed image noise removal based on non-local self-similar dictionary learning
Poissonian and Gaussian noise is main source which affects the image quality captured by new generation of charge-coupled device.Lots of research interests have been paid on the noise suppression,but these works concentrated on the pure Poissonian or Gaussian noise.The proper noise model will improve the performance of noise removing.As noise and signal would be separated efficiently in sparse domain,the denoising performance will be improved if non-local repetitive structure is employed into denoising algorithm based on sparse coding.In this paper we propose a method for Poisson-Gaussian mixed noise reduction by exploiting the sparsity of signal over redundant dictionary and the self-similarity of imagery.Experiments are carried to demonstrate the efficiency and performance of our method over the state-of-art denoising method on a large number of imagerys.

Poissonian and Gaussian noiseImage acquisitionNon-local self-similarDictionary learning

李敏、唐春玲

展开 >

重庆广播电视大学电子信息工程学院,重庆400052

混合噪声 字典学习 非局部 自相似性

Project Supported by Scientific and Technological Research Program of Chongqing Municipal Education CommissionProject Supported by Scientific and Technological Research Program of Chongqing Municipal Education Commission

kj1603810kj1737456

2018

机床与液压
中国机械工程学会 广州机械科学研究院有限公司

机床与液压

CSTPCD北大核心
影响因子：0.32
ISSN：1001-3881
年,卷(期)：2018.46(12)
  • 18