基于改进的K-SVD字典学习的CT图像重建算法

发布时间：2018-05-20 08:02

本文选题：图像重建 + SART算法　；参考：《天津职业技术师范大学》2016年硕士论文

【摘要】：在CT图像重建领域中,重建算法的研究已然取得了傲人的成绩。压缩感知理论的提出为研究不完全角度重建问题提出了新的思路,CS理论只需少量的样本点便能够恢复出原来的信号。而K-SVD字典学习作为图像稀疏的典型算法之一,通过稀疏分解与更新字典,能使稀疏信号与原始图像匹配度提高。本文首先阐述了CT图像重建算法,详细介绍了几种图像重建算法及K-SVD算法,并给出了其仿真实验效果与客观评价分析。其次,鉴于通过K-SVD算法训练得到的自适应字典构造也与初始化字典的选取有关,为此本文通过引进了一个新的初始化字典矩阵,以使K-SVD算法能有更好的图像表达能力,从而与图像重建算法更好的契合。最后,考虑到图像块的结构相似性,将结构相似的图像块通过NLM算法将其聚类,分别对每一类图像块进行字典学习。将改进的K-SVD算法与SART算法结合,进行仿真实验,由具体的评价指标表明,本文算法效果更好。
[Abstract]:In the field of CT image reconstruction, the research of reconstruction algorithm has made a proud achievement. The theory of compressed perception provides a new idea for the study of incomplete angle reconstruction. CS theory can recover the original signal with only a small number of sample points. As one of the typical algorithms of image sparsity, K-SVD dictionary learning can improve the matching degree between sparse signal and original image by sparse decomposition and updating dictionary. In this paper, the CT image reconstruction algorithm is introduced, several image reconstruction algorithms and K-SVD algorithm are introduced in detail, and the simulation results and objective evaluation analysis are given. Secondly, in view of the fact that the construction of adaptive dictionary trained by K-SVD algorithm is also related to the selection of initialization dictionary, a new initialized dictionary matrix is introduced in this paper, so that the K-SVD algorithm can have better image expression ability. Thus, it fits well with the image reconstruction algorithm. Finally, considering the structural similarity of image blocks, the image blocks with similar structure are clustered by NLM algorithm, and each image block is studied by dictionary. The improved K-SVD algorithm is combined with the SART algorithm to carry on the simulation experiment. The concrete evaluation index shows that the algorithm in this paper is more effective.
【学位授予单位】：天津职业技术师范大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【参考文献】