摘要
针对非局部均值滤波的权值由相似块的欧式距来确定而未考虑其受噪声影响的缺点,提出了一种权值由相似块欧式距的统计特性确定的去噪算法。该算法首先对受到高斯噪声干扰的图像相似块的欧式距建立概率分布函数,再由概率分布函数确定权值大小,从而有效地减小高斯噪声对加权系数的影响,以提高去噪性能。实验中,从主客观方面与传统非局部均值滤波进行对比分析,实验数据表明本文提出的算法峰值信噪比提高约1dB,去除噪声的同时保留更多图像的细节信息,去噪性能更优。
In order to overcome the shortcomings that weights of traditional non-local means are only decided by similar block＇s Euclidean distance and the influence of noise is not considered,a novel nonlocal means algorithm based on statistical property of similar block＇s pixels is proposed.This new algorithm first obtains the probability density function of similar block＇s Euclidean distance by employing the prior information of the Gaussian noise,and then formulates the probabilistic weights truly reflecting the similarity between two noisy patches.In the experiments,compared with traditional non-local means algorithm from the subjective and objective,this new algorithm gains 1 dB higher peak signal noise ratio than the traditional non-local means algorithm,and removes the noise more effectively while preserving significant image details.
出处
《液晶与显示》
CAS
CSCD
北大核心
2014年第3期450-454,共5页
Chinese Journal of Liquid Crystals and Displays
基金
人工智能四川省重点实验室开放基金(No.2012RYY08)
四川理工学院校级科研基金(No.2012KY13)
四川省教育厅项目(No.14ZB0211)
四川高校科研创新团队建设计划资助项目(No.13TD0017)
关键词
非局部均值
概率密度
欧式距
图像去噪
non-local means
probability density
euclidean distance
image denoising
