摘要
通过引入局部相位特征,提出了一种新的基于NL-means的低剂量CT图像去噪算法.在非局部滤波器中引入局部相位,设计新的图像子块相似性测度函数,用于低剂量CT图像的去噪.通过与其他4种流行的去噪算法进行模拟图像数值比较,并对真实图像去噪进行临床评价,结果表明:所提出的方法在对低剂量CT图像去除噪声的同时,能保留具有重要诊断价值的CT图像特征,如边界、囊肿区及低密度区等.量化及临床实验结果表明所提出的算法能有效地滤除低剂量CT图像中的噪声并保留图像中有用的诊断信息.
By introducing local phase feature,an enhanced Nonlocal（NL）-means algorithm for noise reduction in low dose computed tomography（LDCT）images was proposed.The similarity measure in NL-means was modified such that neighborhoods with similar phase response receive a larger weight.Compared with four other denoising methods,superior performance was shown in experiments carried out on synthetic images and better restoration of disease features（edges,cystic areas,low-density areas,etc.）was concluded in clinically visual comparisons on real LDCT images.Both quantitative results and clinical evaluation validate good performances of the proposed method in denoising and preserving structural information of LDCT images.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2012年第7期42-46,共5页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金重点资助项目(61031003)
国家自然科学基金资助项目(61031003)
广东省省部产学研结合项目(2011B090400059)
2011年深圳市基础研究项目(JCY20110047)
关键词
CT图像
图像去噪
非局部方法
局部相位
测度函数
computed tomography（CT）images
image denoising
Nonlocal-means
local phase
measure function
