aim-resolve: Automatic Identification and Modeling for Bayesian Radio Interferometric Imaging

TL;DR

aim-resolve introduces an automatic, iterative framework for Bayesian radio interferometric imaging that jointly identifies and models diverse sky components (diffuse background, extended tiles, and point sources) while providing uncertainty quantification. It combines a multi-component sky model with Gaussian-process priors, a Gaussian likelihood, and variational inference (geoVI) via NIFTy, plus fast-resolve for speed, and uses a U-Net plus clustering to detect components and guide model updates. The approach is validated on synthetic data, showing improved reconstruction fidelity and cleaner separation of components compared to single-component methods, and is demonstrated on a MeerKAT L-band observation of ESO 137-006, where it identifies and separates the main galaxies and numerous sources with quantified uncertainties. The work highlights the decoupling of sky emission modeling from instrument response, enabling application across instruments and frequency bands, and outlines pathways for extension to multi-frequency and unified calibration scenarios with practical impact on catalog generation and scientific analyses.

Abstract

Modern radio interferometers deliver large volumes of data containing high-sensitivity sky maps over wide fields-of-view. These large area observations can contain various and superposed structures such as point sources, extended objects, and large-scale diffuse emission. To fully realize the potential of these observations, it is crucial to build appropriate sky emission models which separate and reconstruct the underlying astrophysical components. We introduce aim-resolve, an automatic and iterative method that combines the Bayesian imaging algorithm resolve with deep learning and clustering algorithms in order to jointly solve the reconstruction and source extraction problem. The method identifies and models different astrophysical components in radio observations while providing uncertainty quantification of the results. By using different model descriptions for point sources, extended objects, and diffuse background emission, the method efficiently separates the individual components and improves the overall reconstruction. We demonstrate the effectiveness of this method on synthetic image data containing multiple different sources. We further show the application of aim-resolve to an L-band (856 - 1712 MHz) MeerKAT observation of the radio galaxy ESO 137-006 and other radio galaxies in that environment. We observe a reasonable object identification for both applications, yielding a clean separation of the individual components and precise reconstructions of point sources and extended objects along with detailed uncertainty quantification. In particular, the method enables the creation of catalogs containing source positions and brightnesses and the corresponding uncertainties. The full decoupling of sky emission model and instrument response makes the method applicable to a wide variety of instruments or wavelength bands.

Figures (11)

• Figure 1: Visualization of the iterative method of aim-resolve. Initialized with a single component model (0), it generates a preliminary reconstruction of the data (d). It detects point sources and extended objects in the reconstruction (a) and adds the identified components, marked with crosses and boxes, to the existing background model (b). Next, it separates the components from the background and fits them to the previous reconstructed image (c), before it continues the optimization on the real data (d). This cycle is repeated multiple times for further refinement.
• Figure 2: Comparison of the reconstructions of the synthetic radio data using the single-component model (SCM) and the multi-component model (MCM, after the 3rd iteration of aim-resolve). The first row shows the naturally weighted dirty image of the radio data and the underlying ground truth image. The following rows illustrate posterior mean and relative standard deviation of the reconstructions with the SCM and MCM after full convergence, respectively. The MCM reconstruction shows more precise and detailed reconstructions for the point sources and small extended objects, which is further illustrated by tighter flux contour lines around the sources in the MCM uncertainty map.
• Figure 3: Identified components for several iterations of aim-resolve applied to synthetic image data. The left top image shows the ground truth along with the true components. The remaining images illustrate the reconstructed posterior means for iterations 1 to 3, together with detected point sources (crosses) and extended objects (boxes), showing an increasing number of identified components and an enhancement in overall reconstruction quality across the iterations.
• Figure 4: Separate reconstructions of the components of the MCM in \ref{['fig:exp_recs']}. The left image shows the reconstructed posterior mean of the diffuse background. The right image shows a combined plot of all point sources and extended objects. The sources are well-separated from the background, which captures only the large-scale diffuse emission present in the data.
• Figure 5: Identified components in the 2nd iteration of aim-resolve applied to the ESO 137-006 MeerKAT observation in the LO sub-band (961 - 1145 MHz) with a FOV of $2\degr \times 2\degr$. The two large boxes indicate the identified galaxies ESO 137-006 and ESO 137-007. The remaining detected objects and point sources are marked with small boxes and crosses, respectively.