联合局部和全局稀疏表示的磁共振图像重建方法

发布时间：2018-04-30 17:16

本文选题：压缩感知 + 字典学习　；参考：《重庆大学学报》2017年01期

【摘要】：针对在压缩传感中独立使用全局或局部稀疏字典所分别导致的图像细节或整体图像结构信息的丢失,提出了一种联合利用局部和全局稀疏约束来捕捉磁共振图像细节和整体结构信息的磁共振图像重建算法。该算法首先从特定的磁共振图像中训练出稀疏字典,然后利用该字典进行局部稀疏编码。其次,利用预定义的全局字典来加强磁共振图像的全局稀疏性。最后,在局部和全局稀疏的共同约束下,利用非线性共轭梯度算法来对重建模型进行求解。整个重建过程可以重复迭代以逐步改善重建质量。实验结果表明:当下采样因子达到10时,相比于字典学习算法(dictionary learning MRI,DLMRI),提出的算法在重建质量上可以提高1-6dB。
[Abstract]:For the loss of image details or overall image structure information caused by the independent use of global or local sparse dictionaries in compression sensing, A magnetic resonance image reconstruction algorithm which combines local and global sparse constraints to capture the details of the magnetic resonance image and the global structure information is proposed. The algorithm firstly trains the sparse dictionary from a specific magnetic resonance image and then uses the dictionary for local sparse coding. Secondly, the predefined global dictionary is used to enhance the global sparsity of Mr images. Finally, the nonlinear conjugate gradient algorithm is used to solve the reconstruction model under the constraint of local and global sparsity. The whole reconstruction process can be iterated repeatedly to improve the reconstruction quality step by step. The experimental results show that the current sampling factor reaches 10:00, which can improve the reconstruction quality of 1-6 dB compared with dictionary learning algorithm (dictionary learning / DLMRI).
【作者单位】：重庆大学软件学院;
【基金】：国家自然科学基金青年基金(61402062) 中央高校基本科研业务费专项基金(CDJZR12090003) 重庆市前沿与应用基础研究资助项目(CSTC2015JCYJA40037;CSTC2013JCYJA40038)~~
【分类号】：R445.2;TP391.41

【参考文献】