北京生物医学工程

基于双层字典学习的低剂量CT图像重建算法

Low dose CT image reconstruction algorithm based on thedouble layer dictionary learning method

作者：朱雪茹李勇明李传明李志超王健刘燕

单位：第三军医大学西南医院放射科（重庆400038）2重庆大学通信工程学院（重庆400044）

分类号：R318.04

出版年·卷·期（页码）：2017·36·6（584-590）

摘要：

目的低剂量投影条件下的CT图像重建。方法采用双层K-奇异值分解(K- singular value decomposition，K-SVD)字典训练的学习方法进行图像的超分辨率重建。字典学习方法中采用K-SVD算法，稀疏编码采用正交匹配追踪(orthogonal matching pursuit，OMP)算法。该算法首先利用训练库进行第一层字典训练，然后利用第一层训练的字典对低分辨率图像进行重建。进而将重建图像作为第二层待重建图像的输入，这样使得第二层输入图像含有较多的高频细节信息，因此能在重构的过程中恢复更多的细节信息，让高分辨率重构图像达到较好的效果。结果双层字典重建效果明显优于K-SVD算法，重建图像更接近于原始高分辨率CT图像。结论本研究对双层字典训练学习的框架进行反迭代投影的全局优化改进，改善了图像的重建质量。

Objective Reconstruction of CT image under low dose projection.Methods In this thesis，we adopt double dictionary training method based on K-singular value decomposition(K-SVD)algorithm for super resolution reconstruction of images.In the dictionary learning method，the K-SVD algorithm is adopted，and the sparse coding is orthogonal matching pursuit(OMP)algorithm.Firstly，we use training library to train the first layer dictionary，and then based on the first layer trained dictionary reconstruct low resolution images.Secondly，we put the reconstructed images as the input of the second layer images to be constructed，which make the second layer input image with more high frequency information and restore details in the process of reconstruction.Results The double-layer dictionary reconstruction is superior to the K-SVD algorithm，and the reconstructed image is closer to the original high-resolution CT image.Conclusions In this paper，the global optimization for the inverse iterative projection of the double layer dictionary training is improved，and the quality of the reconstruction image is also improved.

参考文献：

［1］庄天弋.CT原理与算法［M］.上海:上海交通大学出版社，1992:2-93.

Zhuang TY.Principle and algorithm of CT［M］.Shanghai:Shanghai Jiao Tong University Press，1992:2-93.

［2］刘楠.稀疏角度及低剂量CT图像重建算法研究［D］.广州:南方医科大学，2011.

Liu N.Image reconstruction for sparse angular and low-dose computed tomography［D］.Guangzhou:Southern Medical University，2011.

［3］宋慧慧.基于稀疏表示的图像超分辨率重建算法研究［D］.合肥:中国科学技术大学，2011.

Song HH.Research for image super-resolution reconstruction algorithm based on sparse representation［D］.Hefei:University of Science and Technology of China，2011.

［4］Yang JC，Wright J，Huang TS，et al.Image super-resolution as sparse representation of raw image patches［C］//IEEE Conference on Computer Vision and Pattern Recognition.Anchorage，AK:IEEE,2008:1-8.

［5］Yang JC，Wright J，Huang TS，et al.Image super-resolution via sparse representation［J］.IEEE Transactions on Image Processing，2010，19(11):2861-2873.

［6］Aharon M，Elad M，Bruckstein A.K-SVD:an algorithm for designing overcomplete dictionaries for sparse representation［J］.IEEE Transactions on Signal Processing，2006，54(11):4311-4322.

［7］齐宏亮.稀疏角度下的CT图像重建迭代算法研究［D］.广州:南方医科大学，2013.

Qi HL.Research on the CT image reconstruction from sparse view data using iterative methods［D］.Guangzhou:Southern Medical University，2013.

［8］魏艳新.基于字典训练和稀疏表示的图像超分辨率重建算法研究［D］.秦皇岛:燕山大学，2013.

Wei YX.Super-resolution reconstruction based on dictionary training and sparse representation［D］.Qinhuangdao:Yanshan University，2013.

［9］Zeyde R，Elad M，Protter M.On single image scale-up using sparse-representations［C］//International Conference on Curves and Surfaces. Berlin:Springer,2010,6920:711-730.

［10］Freeman WT，Pasztor EC.Learning low-level vision［J］.International Journal of Computer Vision，2000，40(1):25-47.

［11］Sun J，Zheng NN，Tao H，et al.Image hallucination with primal sketch priors［C］//IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Madison，Wisconsin:IEEE,2003,2:729-736.

［12］Zhang J，Zhao C，Xiong RQ，et al.Image super-resolution via dual-dictionary learning and sparse representation［C］//IEEE International Symposium on Circuits and Systems. Seoal:IEEE，2012:1688-1691.

［13］马小薇.基于压缩感知的OMP图像重构算法改进［J］.电子科技，2015，28(4):51-53.

Ma XW.Improvement of OMP image reconstruction algorithm based on compressed sensing［J］.Electronic Science and Technology，2015，28(4):51-53.

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录