First AI for Deep Super-resolution Wide-field Imaging in Radio Astronomy: Unveiling Structure in ESO 137-006

Abstract

We introduce the first AI-based framework for deep, super-resolution, wide-field radio interferometric imaging and demonstrate it on observations of the ESO 137-006 radio galaxy. The algorithmic framework to solve the inverse problem for image reconstruction builds on a recent “plug-and-play” scheme whereby a denoising operator is injected as an image regularizer in an optimization algorithm, which alternates until convergence between denoising steps and gradient-descent data fidelity steps. We investigate handcrafted and learned variants of high-resolution, high dynamic range denoisers. We propose a parallel algorithm implementation relying on automated decompositions of the image into facets and the measurement operator into sparse low-dimensional blocks, enabling scalability to large data and image dimensions. We validate our framework for image formation at a wide field of view containing ESO 137-006 from 19 GB of MeerKAT data at 1053 and 1399 MHz. The recovered maps exhibit significantly more resolution and dynamic range than CLEAN, revealing collimated synchrotron threads close to the galactic core.

Scalable precision wide-field imaging in radio interferometry: II. AIRI validated on ASKAP data

“…In this section, we briefly recall the RI data model in the context of wide-field imaging and provide a summary of AIRI, building from the underlying theory (Terris et al 2022) and its first application to real RI data (Dabbech et al 2022). We also outline the encompassing framework for wide-field imaging, focusing on the parallelisation of the AI denoiser and the automated selection of associated parameters.…”

“…Implemented in MATLAB, the framework is automated, highly parallelised, and capable of producing wide-field, high-dynamic range, super-resolved monochromatic intensity images. Two interchangeable image regularisation denoisers can be "plugged" in as the 'backward' step of the underlying iterative forward-backward (FB) deconvolution structure (Terris et al 2022;Dabbech et al 2022) of the imaging framework, alternating with a gradient descent 'forward' step promoting data fidelity.…”

“…The first algorithm, unconstrained Sparsity Averaging Reweighted Analysis (uSARA), is purely optimisation-based. It leverages the proximal operator of a state-of-the-art handcrafted sparsitypromoting regularisation function (Carrillo et al 2012;Terris et al 2022) as regularisation denoiser. In Part I: "uSARA validated on ASKAP data", uSARA was validated against the widely-used CLEAN-based imager for RI WSClean (Offringa et al 2014;Offringa & Smirnov 2017).…”

“…In this sequel, we showcase AIRI (Terris et al 2022) -the second algorithm encapsulated in our automated, parallelised imaging framework -which combines optimisation theory with the power of AI. Building on the recent success of Plug-and-Play (PnP) approaches in various applications such as image restoration (Zhang et al 2021) and magnetic resonance imaging (Ahmad et al 2020), AIRI relies on the same FB iterative scheme as uSARA, with the proximal operator enforcing the image prior model replaced by a learned DNN denoiser.…”

High performance w-stacking for imaging radio astronomy data: a parallel and accelerated solution

“…Also of note are comprehensive and widely adopted software platforms, able to perform most of the previous tasks, like CASA (McMullin et al 2007), AIPS (Greisen 2003), MIRIAD (Sault et al 1995) and ASKAPsoft (Wieringa et al 2020). Finally, it is worth mentioning the two first applications of Deep Learning for imaging presented by Terris et al (2022) and Dabbech et al (2022).…”