Magnetic Resonance Image Quality Metrics Based on Sparse Dictionary Learning

Assessing the quality of Magnetic Resonance (MR) images is an important diagnostic tool in clinical settings. This paper describes an approach based on sparse dictionary learning (SDL) that reliably quantifies the image quality of MR data. The proposed metric is both reference-free, i. e., does not require a high-quality image of the same target, and agnostic to both the measurement and the reconstruction protocols.