Jinhong Huang1,2, Biaoshui Liu1, Gaohang Yu1,2, Yanqiu Feng1, and Wufan Chen1
1School of Biomedical Engineering and Guangdong Provincial Key Laboratory of Medical Image Processing, South Medical University, Guangzhou, China, People's Republic of, 2School of Mathematics and Computer Science, Gannan Normal University, Ganzhou, China, People's Republic of
Conventional
CS methods treat a 2D/3D image to be reconstructed as a vector. However, many
data types do not lend themselves to vector data representation, and this
vectorization based model may lose the inherent spatial structure property of
original data and suffer from curse of dimensionality that occurs when working
with high-dimensional data. In this work, we introduce a novel tensor
dictionary learning method for dynamic MRI reconstruction. Numerical experiments
on synthetic data and in vivo data show approximately 2 dB improvement in PSNR
presented by the proposed scheme over existing method with overcomplete
dictionary learning.
