多图像引导滤波器去除MRI Rician噪声技术

90 浏览量更新于2024-08-28 收藏 455KB PDF 举报

"这篇研究论文提出了一种使用多图像引导滤波器消除磁共振成像（MRI）中的Rician噪声的新方法。该方法由作者Long Chen、Junwei Duang和C.L. Philip Chen在澳门大学计算机与信息科学系共同完成。论文通过分析Rician噪声的特性，提出了一种两步去噪方案，包括偏差减少步骤和去噪步骤。在去噪步骤中，设计了一种使用预处理图像作为指导的滤波器，以提高去噪效果。实验结果证明了所提方法在Rician噪声去除问题上的优越性能。" 在医学成像领域，磁共振成像（MRI）是一种重要的诊断工具，它能提供高分辨率的软组织图像。然而，MRI图像往往会受到各种类型的噪声干扰，其中Rician噪声是最常见的一种。Rician噪声是由MRI信号的非高斯分布引起的，导致图像出现模糊和失真，影响医生对图像的分析和诊断。本文提出的多图像引导滤波器（Multi-image Guided Filter, MGF）是一种创新的噪声去除策略。它考虑到Rician噪声的特点，即噪声成分在图像的高频和低频部分都有分布，因此传统的单图像滤波器可能无法有效地去除这种噪声。MGF采用两步法来处理这个问题： 1. 偏差减少步骤：这一步旨在减小噪声对图像结构的影响，特别是对图像边缘和细节的破坏。通过某种预处理方式，可以部分地分离出噪声和图像信号，从而降低后续处理的难度。 2. 去噪步骤：这是关键的一步，多图像引导滤波器在此发挥作用。该滤波器利用多个经过预处理的图像作为指导，这些图像可能包含了不同的噪声信息或不同的图像特征。通过比较和融合这些图像，滤波器能够更准确地识别和去除噪声，同时保持图像的结构完整性。实验部分，作者对比了所提方法与其他常见的去噪算法，如基于快速傅里叶变换的滤波器、非局部均值去噪等。结果显示，MGF在保留图像细节和结构的同时，能够更有效地消除Rician噪声，提高了图像的质量和诊断价值。这项工作对于MRI图像的后处理有着重要的理论和实践意义，它为处理Rician噪声提供了新的思路，并且有望在未来推动医疗成像技术的进步。

Rician Noise Removal in MR Imaging Using

Multi-image Guided Filter

Long Chen

Department of Computer

and Information Science

University of Macau

Taipa,Macau, China

Email: longchen@umac.mo

Junwei Duan

Department of Computer

and Information Science

University of Macau

Taipa,Macau, China

Email: yb27409@umac.mo

C. L. Philip Chen

Department of Computer

and Information Science

University of Macau

Taipa,Macau, China

Email: philipchen@umac.mo

Abstract—Rician noise removal from a given Magnetic Res-

onance image is the denoising problem we address here. This

paper proposes a novel Rician noise removal approach based

on multi-image guided ﬁlter (MGF). Considering the special

characteristics of Rician noise, we propose a two-step scheme

for the purpose of noise removal. In this scheme, bias-reducing

step and denoising step are included. In the denoising step, a

guided ﬁlter is deigned to apply multiple preﬁltered images as

the guidance. The experiments demonstrate that our proposed

method can achieve good performance on the rician noise removal

problem.

Index Terms—Rician noise, MR imaging, multi-image guided

ﬁlter, Image denoising

I. INTRODUCTION

Magnetic resonance imaging (MRI) are widely used in

different medical applications. However, due to the limitation

of current acquisition technology, MRI images are usually

contaminated by different noises and artifacts [11], so that

the acquired images usually contain inaccurate information

and the quality is also unsatisfactory. It’s worse that the

image with low quality will affect further analysis such as

segmentation[4], fusion[3], and pattern recognition. Thus, it’s

necessary to do some preprocess steps like denoising before

further analysis.

In order to know the noise in image better, the additive noise

model, as the most common model, is introduced as follows.

= I

mtrue

+ N

noise

(1)

where, I

mtrue

is the signal without any noise, I

is the

observed signal and N

noise

is the noise. By using this noise

model, a large number of de-noising methods have been

developed. In this paper, we focus on Rician noise in MRI

images. In the MRI’s complex raw data, the real and imaginary

parts are corrupted by white additive Gaussian noise. And

the noise variance is assumed to be the same in both parts

[14]. The noise can be modeled as the Rician noise if taking

the magnitude of the complex data. In other words, the

noise in magnitude of MRI images usually follows the Rician

distribution. The removal of Ricain noise is more complicated

than the classical noises such as Gaussian noise.

Generally speaking, two typical ways are adopted to reduce

the noise in MRI images. One way is to average the data

obtained several times. Another way is to reduce the noise

by the post processing methods. Until now, to remove the

noised in MRI images, a large number of post processing

denoising methods, such as ﬁltering approaches, transform

domain approaches and statistical approaches [1], have been

proposed. Although the progress is huge, obtaining desirable

denoising results is still a challenge problem [8]

In the current denoising methods, spatial ﬁltering methods

and transform domain ﬁltering methods are the two major

categories. In an image, as we know, it directly deals with

pixels for spatial ﬁltering methods. For a spatial ﬁlter, through

a special function of the intensities of the pixels within a

neighborhood, each pixel value I(u) can be transformed.

Traditional spatial image ﬁlters include Gaussian ﬁlter [9],

median ﬁlter[2], bilateral ﬁlter and Wiener ﬁlter[12]. In a

region with small size, Gaussian ﬁlter and median ﬁlter can

remove noise and blur images. Most of these methods can

signiﬁcantly remove the noise in the image, however, they

blur and add artifacts to the images. In the transform domain

ﬁltering methods, the ﬁlter ﬁrst changes the images from space

domain into some wavelet domain or other transform domains

[7] and then processes the images in the domain we changed

into. Similar to the spatial ﬁltering methods, the transform

domain ﬁltering methods can also reduce the noise, meanwhile

they also generate some artifacts.

In this paper, the key point we mainly focus on is the

ﬁlter in spatial domain. As a spatial domain ﬁlter, the recently

proposed guided ﬁlter has demonstrated its better performance

of edge-preserving [10]. Mathematically, this ﬁlter is a local

linear model between guidance image and output image after

ﬁltering. However, it usually can only use one image to guide

the process. Thus, we extend the guided ﬁlter using the several

images. In other words, we can utilize the information from

many images to do the image processing, which is one of the

contributions of the multi-image guided ﬁlter introduced here.

Multi-image guided ﬁlter will described in detail in section

3. Before using multi-image guided ﬁlter to remove the noise,

we adopt the variance-stabilization transform (VST) to change

the rician noise to the Gaussian one. [6]

The rest of this paper is organized as follows. A brief survey

of the related works and methods are present in Section II.

180

2016 3rd International Conference on Informative and Cybernetics for Computational Social

stems (ICCSS)

下载后可阅读完整内容，剩余4页未读，立即下载

weixin_38664532

粉丝: 9
资源: 945

多图像引导滤波器去除MRI Rician噪声技术

使用Presmooth非局部均值滤波器对Rician噪声进行脑MR图像去噪

图像处理MATLAB的噪声图像高斯噪声.docx

使用预平滑非局部均值滤波器对Rician噪声进行脑MR图像去噪

通信消除杂音MatLab程序

matlab Rician衰落

图像处理运用python分别利用逆滤波、维纳滤波、约束最小二乘方滤波方法对模糊图像2名为Motion Blurred Image进行滤波处理，得到图像3，图像4，图像5，分别显示出图像3，图像4，图像5。

rician衰落分布莱斯

Rician 信道模型

调制识别数据集RML2016.10a和RML2016.10b中，数据集中包含的Rician信道和Rayleigh信道之间的差异是什么

怎么用matlab完成信道建模

最新资源