首页|基于BM3D的脑MRI图像噪点剔除算法

基于BM3D的脑MRI图像噪点剔除算法

扫码查看
原文链接
  • 国家科技期刊平台
  • NETL
  • NSTL
  • 万方数据
磁共振成像(Magnetic Resonance Imaging，MRI)已经成为一种常见的影像检查方式，MRI的去噪算法影响着MRI的成像效果。基于深度学习的MRI去噪算法需要一定量的数据，绝大部分基于非深度学习的MRI去噪算法都是将MRI数据转化为实数之后进行去噪的，针对复数MRI中的复数数据类型的算法也存在着失真的问题。因此，提出一种通过单张MRI脑图像的原始数据进行噪点剔除的算法，以此更好得去除图像噪声。该算法从MRI的原始数据出发，利用了MRI噪声分布性质和MRI脑图像的特点，以判断MRI图像中噪声明显的点，从而剔除MRI中特定的莱斯分布的噪声。并将所提出的算法结合了MRI 去噪中常用的非局部平均算法(Non-Local Means denoising，NLM)与三维块匹配算法(Block-Matching and 3D filtering，BM3D)，并和不使用该算法剔除噪点的NLM、BM3D进行了对比评估。对比结果表明，在噪声密度不同的多种情况下，该算法总能优化与之相结合的图像去噪算法，在不同的噪声情况下使峰值信噪比(Peak Signal-to-Noise Ratio，PSNR)与结构相似性(Structural Similarity，SSIM)提高了1%～9%。最后将该算法结合BM3D，对比了DnCNN、低秩聚类算法(Weighted Nuclear Norm Minimization，WNNM)、BM3D、NLM等用于MRI去噪的算法，在莱斯噪声较多时，该算法在PSNR上有更好的表现。
Complex Domain Noise Removal Algorithm for a Single Brain MRI Image Based on BM3D
Magnetic resonance imaging(MRI)has become a common imaging examination method,and the denoising algorithm of MRI affects the imaging effect.Deep learning based MRI denoising algorithms require a certain amount of data,and the vast majority of non deep learning based MRI denoising algorithms convert MRI data into real numbers for denoising.Algorithms for complex data types in complex MRI data also have distortion issues.Therefore,a noise removal algorithm is proposed based on the raw data of a single MRI brain image to better remove image noise.Starting from the raw data of MRI,the proposed algorithm utilizes the distribution properties of MRI noise and the characteristics of MRI brain images to determine the points with obvious noise in the MRI image,and thus eliminates specific Rician distribution noise in the MRI.And the proposed algorithm was combined with the commonly used Non-Local Means de-noising(NLM)and Block-Matching and3D filtering(BM3D)in MRI denoising,and the denoising effect was compared and evaluated with NLM and BM3D denoising algorithms that did not use this algorithm to remove noise.The comparison results show that in various situations with different noise densities,the proposed algorithm can always optimize the image denoising algorithm combined with it,and improve Peak Signal-to-Noise Ratio(PSNR)and Structural Similarity(SSIM)by 1%to 9%under different noise conditions.Finally,it was combined with BM3D and compared with other MRI denoising algorithms such as DnCNN,Weighted Nuclear Norm Mini-mization(WNNM),BM3D,and NLM.When there is more Rician noise,the proposed algorithm performs better on PSNR.

brain MRI imageingnoise removalRician distributionNon-Local Means denoising(NLM)Block-Matching and 3D filtering(BM3D)

徐梦笔、何刚

展开 >

西南科技大学 计算机科学与技术学院,四川 绵阳 621010

国家卫生健康委核技术医学转化实验室(绵阳市中心医院),四川 绵阳 621010

脑磁共振成像 噪声去除 莱斯分布 非局部平均算法 三维块匹配算法

四川省科技项目四川省科技项目国家卫生健康委员会核技术医学转化重点实验室开放课题资助

2020YFS04542020YFS03182021HYX031

2024

10.20165/j.cnki.ISSN1673-629X.2024.0171

计算机技术与发展
陕西省计算机学会

计算机技术与发展

CSTPCD
影响因子：0.621
ISSN：1673-629X
年,卷(期)：2024.34(9)