Plug-and-Play ADMM for Quantitative Parameter Mapping: A Comparison of Compressive Sensing Algorithms
When
Inversion problems are generally cast as constrained optimizations including a forward (data fidelity) component and a regularization (prior) component. Alternating direction method of multipliers (ADMM) yields a variable splitting scheme to separate the forward and regularization components, and thus it enables a prior or regularizer to be plugged in inversion independently. In this talk, we will review three journal papers that focus on Magnetic Resonance (MR) parameter mapping using model based Compressive Sensing (CS) approaches. We will illustrate that ADMM is suited for solving such complex physical sensor model (e.g. Magnetic Resonance Imaging) with any proper regularity criteria.