基于字典学习方法的CT不完全投影图像重建算法  

Incomplete Project Data Image Reconstruction Algorithm Based on Dictionary Learning Method

作　　者：赵可[1] 潘晋孝[1] 孔慧华[1] 

机构地区：[1]中北大学电子测试技术国家重点实验室,山西太原030051

出　　处：《数学的实践与认识》2014年第2期143-149,共7页Mathematics in Practice and Theory

基　　金：973计划(2011CB311804);国家自然科学基金(61071193;61171179;61227003;11247244);山西省自然科学基金(2010011002-1;2010011002-2;2012021011-2);高等学校博士学科点专项科研基金(20121420110006);山西省高等学校优秀创新团队支持计划资助

摘　　要：对于不完全投影角度的重建研究是CT图像重建中一个重要的问题.将压缩感知中字典学习的方法与CT重建算法ART迭代算法相结合.字典学习方法中字典更新采用K-SVD(K-奇异值分解)算法,稀疏编码采用OMP(正交匹配追踪)算法.最后通过对标准Head头部模型进行仿真实验,验证了字典学习方法在CT图像重建中对于提高图像的重建质量和提高信噪比的可行性与有效性.For the problem of incomplete projection data of CT image reconstruction,CS Theory attracts more and more attention in recent years.The dictionary learning method in CS Theory is applied to image reconstruction algorithm such as ART iterative algorithm,which effectively improves the quality of image reconstruction.Dictionary updates using KSVD algorithm and sparse coding using OMP algorithm in dictionary learning method.In this paper,we reconstruct the Head standard model with the ART algorithm and the algorithm in this paper proposed.The simulation experimental results show that the feasibility and effectiveness dictionary learning methods for improving the quality of image reconstruction and noise suppression in incomplete projection data of CT image reconstruction.

关 键 词：不完全投影 压缩感知 字典学习 K-SVD算法 OMP算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]