摘要
提出一种利用组稀疏表示进行CSMRI的方法.在字典学习过程中,对图像块按照相似性准则进行分组,并利用这些组进行字典训练.将组字典学习的代价函数引入到压缩感知核磁共振成像的模型中,并利用交替优化方法求解该模型.提出的算法不仅利用了图像的局部稀疏性,还利用了图像块之间的相似性(非局部相似性).实验结果证明,该算法能够重构出高质量图像.
A method of utilizing group sparse representation for compressed sensing magnetic resonance imaging (CSMRI) is proposed. In the dictionary learning process, the image patches are grouped based on the similarity metric, and these groups are utilized for dictionary training. The cost function of group dictionary learning is incorporated into the model of compressed sensing magnetic resonance imaging, and alternating optimization method is proposed to solve the corresponding problem. Both the local sparsity of the image and the similarity between image patches (non-local similarity) are utilized in the proposed algorithm. The experimental results indicate that the algorithm can reconstruct images with high quality.
出处
《内蒙古师范大学学报(自然科学汉文版)》
CAS
北大核心
2017年第1期90-97,101,共9页
Journal of Inner Mongolia Normal University(Natural Science Edition)
基金
山西省重点实验室开放课题(2016002)
忻州师范学院重点学科建设项目(XK201504)
关键词
压缩感知
核磁共振成像
组稀疏表示
字典学习
非局部相似性
compressed sensing
magnetic resonance imaging(MRI)
group sparse representation
dictionary learning
nonlocal similarity