摘要
在含噪图像的二维经验模态分解(BEMD)的基础上,从图像BEMD分解系数的统计特性出发,构造图像BEMD系数的概率密度函数模型,提出了一种基于相邻尺度间BEMD系数相关性的图像消噪方法,消噪的过程中同时考虑本层BEMD系数特性以及其父层BEMD系数的值。从而能更好地消除噪声,同时更有效地保留图像边缘、纹理等细节信息。实验结果表明,与经典的小波阈值消噪和BEMD阈值消噪算法相比,经本文方法消噪后图像质量有较好的提高,具有更低的均方误差和更高的峰值信噪比。
After denoised image is decomposed by bidimensional empirical mode decomposition（ BEMD）, the probability density function（PDF） is constructed by the statistical properties of BEMD coefficients. A new de- noising method is proposed based on the coherence of adjacent level BEMD coefficients and PDF of BEMD co- efficients. The experimental results show that the proposed method can remove the noise better, and preserve the details of edges and textures more effectively. Compared with the wavelet threshold denoising and BEMD threshold denoising, the images denoised by the proposed method have lower mean square error, higher peak signal to noise ratio and better visual effects.
出处
《测控技术》
CSCD
2015年第6期24-26,30,共4页
Measurement & Control Technology
基金
国家自然科学基金资助项目(41071270
11201354)
关键词
二维经验模态分解
图像消噪
概率密度函数
系数相关性
bidimensional empirical mode decomposition
image denoising
probability density tunction
coetti-cient correlation
