基于多相位图像相似性和全局图割的肺4D-CT超分辨率重建研究
发布时间：2019-01-27 12:42
【摘要】：肺4D-CT图像为放射治疗提供了全面的图像引导,它可以清楚地反映肺部靶区器官随呼吸的运动规律,同时减少呼吸运动引起的伪影,为当今的肺癌治疗提供了巨大的帮助。然而,肺4D-CT数据的采集时间长,剂量大,因此不能实现薄层厚的扫描,这导致采集到的肺4D-CT图像层间分辨率显著低于层内分辨率。因此想要观察正常分辨率的肺4D-CT的Z轴图像,必须要对的Z轴图像按照一定比例进行插值。常用的插值方法有线性插值,三次样条插值等,但是这些插值的方法都不能得到理想的重建结果。本文为了提高肺4D-CT图像的多平面显示的质量,研究了基于图像超分辨率重建的方法,来重建高分辨率的肺4D-CT图像。图像超分辨率重建技术的基本思想是利用单幅或者多幅图像的信息来提高图像的显示质量。常用的方法有基于频域的方法和基于空域的方法。基于频域的方法不易扩展,且不易添加约束。因此目前的研究主要集中在基于空域的超分辨率重建方法研究。本文针对肺4D-CT图像特性,研究了两种基于空间域处理的肺4D-CT图像超分辨率重建:基于多相位图像相似性和基于全局图割的超分辨率重建方法。而且重建结果在视觉和量化方面,均优于传统的插值算法和POCS算法。本文的前两章分别介绍了研究意义和相关技术现状。在第三章和第四章分别详细介绍了基于多相位相似性和基于全局图割的超分辨率重建方法。其中第三章阐述了非局部均值滤波,基于多相位相似性的图像块搜素,以及全局约束重建。第四章阐述了全局图的构造,能量函数构建及优化求解。本文第五章对全文进行了总结和展望。
[Abstract]:Lung 4D-CT images provide a comprehensive image guide for radiotherapy, which can clearly reflect the movement of lung target organs with respiration, and reduce the artifacts caused by respiratory movements, which provides a great help for the treatment of lung cancer today. However, the collection time of lung 4D-CT data is long and the dose is large, so it is not possible to realize thin-layer thick scanning, which leads to the interlayer resolution of collected lung 4D-CT images being significantly lower than that of intralayer resolution. Therefore, in order to observe Z-axis images of lung 4D-CT with normal resolution, the Z-axis images must be interpolated in a certain proportion. The commonly used interpolation methods include linear interpolation, cubic spline interpolation and so on, but none of these interpolation methods can obtain ideal reconstruction results. In order to improve the quality of multiplanar display of lung 4D-CT images, a method based on super-resolution reconstruction is studied to reconstruct high-resolution lung 4D-CT images. The basic idea of image super-resolution reconstruction is to improve the display quality of images by using the information of single or multiple images. The commonly used methods are frequency-based and spatial-based methods. The method based on frequency domain is not easy to extend, and it is difficult to add constraints. Therefore, the present research focuses on spatial-based super-resolution reconstruction. According to the characteristics of lung 4D-CT images, two super-resolution reconstruction methods based on spatial domain processing for lung 4D-CT images are studied in this paper: one is based on the similarity of multi-phase images and the other is based on global image cutting. Moreover, the reconstruction results are superior to the traditional interpolation algorithm and POCS algorithm in vision and quantization. The first two chapters of this paper respectively introduce the significance of the research and the current situation of related technologies. In chapter 3 and chapter 4, the super-resolution reconstruction method based on multi-phase similarity and global graph cut is introduced in detail. In the third chapter, non-local mean filter, image block search based on multi-phase similarity, and global constrained reconstruction are discussed. In chapter 4, the construction of global graph, the construction of energy function and the optimization solution are discussed. The fifth chapter summarizes and prospects the full text.
【学位授予单位】：南方医科大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R734.2;TP391.41
本文编号：2416267
[Abstract]:Lung 4D-CT images provide a comprehensive image guide for radiotherapy, which can clearly reflect the movement of lung target organs with respiration, and reduce the artifacts caused by respiratory movements, which provides a great help for the treatment of lung cancer today. However, the collection time of lung 4D-CT data is long and the dose is large, so it is not possible to realize thin-layer thick scanning, which leads to the interlayer resolution of collected lung 4D-CT images being significantly lower than that of intralayer resolution. Therefore, in order to observe Z-axis images of lung 4D-CT with normal resolution, the Z-axis images must be interpolated in a certain proportion. The commonly used interpolation methods include linear interpolation, cubic spline interpolation and so on, but none of these interpolation methods can obtain ideal reconstruction results. In order to improve the quality of multiplanar display of lung 4D-CT images, a method based on super-resolution reconstruction is studied to reconstruct high-resolution lung 4D-CT images. The basic idea of image super-resolution reconstruction is to improve the display quality of images by using the information of single or multiple images. The commonly used methods are frequency-based and spatial-based methods. The method based on frequency domain is not easy to extend, and it is difficult to add constraints. Therefore, the present research focuses on spatial-based super-resolution reconstruction. According to the characteristics of lung 4D-CT images, two super-resolution reconstruction methods based on spatial domain processing for lung 4D-CT images are studied in this paper: one is based on the similarity of multi-phase images and the other is based on global image cutting. Moreover, the reconstruction results are superior to the traditional interpolation algorithm and POCS algorithm in vision and quantization. The first two chapters of this paper respectively introduce the significance of the research and the current situation of related technologies. In chapter 3 and chapter 4, the super-resolution reconstruction method based on multi-phase similarity and global graph cut is introduced in detail. In the third chapter, non-local mean filter, image block search based on multi-phase similarity, and global constrained reconstruction are discussed. In chapter 4, the construction of global graph, the construction of energy function and the optimization solution are discussed. The fifth chapter summarizes and prospects the full text.
【学位授予单位】：南方医科大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R734.2;TP391.41
【参考文献】
相关期刊论文 前7条
1 吴秀秀;肖珊;张煜;;基于配准的肺4D-CT图像超分辨率重建[J];电子学报;2015年02期
2 王帅;;图像插值算法综述及分析[J];华章;2014年07期
3 苏衡;周杰;张志浩;;超分辨率图像重建方法综述[J];自动化学报;2013年08期
4 张洪艳;沈焕锋;张良培;李平湘;袁强强;;基于最大后验估计的影像盲超分辨率重建方法[J];计算机应用;2011年05期
5 徐美芳;刘晶红;;基于边缘保持的航拍图像凸集投影超分辨率重建算法[J];液晶与显示;2010年06期
6 徐大宏;王润生;;基于非局部正则化的图像去噪[J];计算机应用研究;2009年12期
7 张书旭;周凌宏;陈光杰;林生趣;沈国辉;;4D-CT重建及其在肺癌放疗中的应用研究进展[J];中国辐射卫生;2008年03期
相关硕士学位论文 前1条
1 张锐娟;图像配准理论及算法研究[D];西安电子科技大学;2009年
，本文编号：2416267
本文链接：https://www.wllwen.com/yixuelunwen/zlx/2416267.html
