MRI Reconstruction with Joint Global Regularization and Transform Learning

Reconstruction for Magnetic Resonance Imaging (MRI) is an important inverse problem in biomedical image processing. Successful reconstruction of the magnetic resonance (MR) image from heavily undersampled k-space Fourier samples necessitates the employment of advanced regularization techniques. In the last decade, a primary resource for regularization has been the quest for sparsity in a transform domain. As an example, discrete total variation (TV) minimization (Chambolle, 2004) and wavelet transform sparsity based regularization (Figueiredo and Nowak, 2003) have been widely utilized in various ill-conditioned image restoration problems.
Source: Computerized Medical Imaging and Graphics - Category: Radiology Authors: Source Type: research