摘要
为了更好地利用图像先验以及保护图像边缘、纹理等细节信息,提出一种结合反应扩散(Trained Nonlinear Reaction Diffusion,TNRD)与基于块组先验去噪(Patch Group Prior based Denoising,PGPD)的改进算法。首先,对PGPD去噪后的图像进行小波分解得到3个正交的子带,由理论分析可知图像为各子带之和;然后利用反应扩散对高频系数大于阈值的子带部分进行扩散处理,并将处理结果替代原来部分从而获得最终去噪图像。实验结果表明,改进算法在峰值信噪比、保护细节信息等方面都有较大的性能改善。
In order to make better use of image prior and to protect the edges,texture of the image,an improved algorithm which combines trained nonlinear reaction diffusion（TNRD） with patch group prior based denoising（PGPD） is proposed. First of all,three orthogonal subbands are obtained by wavelet decomposition of the PGPD denoised image. The theoretical analysis shows that the image is the sum of the subbands. Then diffusing the subband with high frequency coefficients of which the wavelet coefficients are greater than the threshold by nonlinear reaction diffusion,and the results are used to replace the original subband to get the final denoising image. The experimental results show that the improved algorithm has greater performance improvement,such as the Peak Signal to Noise Ratio（PSNR）,the details protection.
作者
莫佩基
雷宏
MO Pei-ji;LEI Hong(Institute of Electronics,Chinese Academy of Sciences,Beijing 100190,China;University of Chinese Academy of Sciences,Beijing 100049,China)
出处
《计算机与现代化》
2018年第6期54-57,63,共5页
Computer and Modernization
关键词
图像去噪
反应扩散
块组先验去噪
小波分解
image denoising
reaction diffusion
patch group prior based denoising
wavelet decomposition
