【作者】 黄世亮；

【导师】 叶朝辉； 裘鉴卿；

【作者基本信息】 中国科学院研究生院（武汉物理与数学研究所） ， 无线电物理， 2006， 博士

【摘要】 磁共振成像以其对人体无损害、可以采用多种参数成像和能够反映器官或组织的生化特征等特点，成为科研和医学临床诊断的重要手段之一。本论文围绕磁共振成像中的几个突出问题：磁共振图像截断伪影去除、磁共振图像去噪和磁共振图像与其它模态医学图像的融合为主题，开展一系列的研究工作。 减少相位编码数量导致的磁共振图像截断伪影只沿图像的一个方向(水平方向)，利用二进小波变换将含截断伪影的图像分解为近似子图和水平及垂直细节子图，充分利用磁共振图像截断伪影是由垂直细节子图中的某些频率分量决定的这一特点，为尽可能地保持图像的细节特征，只对含截断伪影的垂直细节子图进行处理(现有文献都是对所有细节子图进行全局处理，因而损失了图像的细节特征。)而其它子图不受损伤。具体在处理时又采用了三种方案：(1) 利用近似子图和各级水平细节子图构造平滑图像，然后按特定的准则向平滑图像添加高频成分以提高图像的空间分辨率。(2) 只对垂直细节子图进行阈值收缩去除截断伪影，然后进行图像重建。(3) 对垂直细节子图进行多分辨率小波分解，去除导致截断伪影的主要信号分量之后进行图像重建。 小波变换和偏微分方程在图像处理中的应用是近年来图像处理中的两个最新进展。在磁共振图像去噪方面，我们提出了一个基于二进小波变换的直接对复数磁共振图像数据去噪的算法，与传统的Wiener去噪相比该算法具有更好的去噪效果。传统的各向异性扩散去噪是利用图像梯度构造控制扩散滤波的扩散张量，我们基于小波变换提取的图像特征构造扩散张量，取得了更好的去噪效果。 图像融合是图像处理中的关键技术之一。它在军事和民用图像处理领域获得了更多还原

【Abstract】 Magnetic resonance imaging technique has become one of the important tools in scientific research and clinic for its non-invasive, imaging with several parameters and reflecting biochemistry characterizations etc. This dissertation is focused on several important issues in magnetic resonance (MR) image processing. We did some studies on truncation artifact removal, MR image denoising and MR image fusion.Since the truncation artifact appear only in one of the two spatial directions, we decomposed the image with truncation artifact into approximate sub-images and vertical and horizontal detailed sub-images. We found that the truncation artifact caused by reducing the number of phase-coded signals were determined by some frequency components in vertical sub-images. In order to reduce the truncation artifact efficiently and preserve the details of the image as much as possible, we only manipulated operation to the vertical sub-images. The sub-images which had no relation with truncation artifact and the components with less truncation artifact are intact in image reconstruction. In this dissertation , we proposed three schemes for truncation artifact removal. We constructed smooth image with approximate sub-image and all parallel detailed sub-images. Then we reconstructed final image by adding details following a special rule. We only did artifact reduction to the vertical sub-images via wavelet shrinkage, then reconstructed the image with inverse dyadic wavelet transform. We decomposed the vertical sub-images further, retained the components which had less artifact更多还原