SARA

Sparsity Averaging Reweighted Analysis

"It is princess SARA, not Leia!"

The SARA family is a suite of optimisation algorithms propelled by a bespoke handcrafted regularization denoiser: the proximal operator of a log-sum prior generalising the l1-norm, and enforcing strong sparsity simultaneously in multiple wavelet bases. Their primary application is to solve large-scale high-resolution high-dynamic range inverse problems for SII in radio astronomy. Firstly, the SARA algorithm itself is built on a Primal-Dual Forward-Backward optimisation algorithmic backbone enabling handling hard data-fidelity constraints, and addresses 2D planar monochromatic intensity imaging. Polarised SARA provides functionality for 2D planar monochromatic polarisation imaging, while Faceted HyperSARA provides functionality for 3D Euclidean wideband intensity imaging. Secondly, uSARA (standing for "unconstrained SARA") relies on a Forward-Backward optimisation algorithmic backbone supporting soft data-fidelity terms. Its JointSARA variant provides functionality for joint calibration and imaging.

As of early 2024, the transfer of uSARA to MRI and ultrasound imaging in medicine is the subject of active developments at BASP. Stay tuned!

Papers & Codes