# An optimized algorithm for multi-scale wideband deconvolution of radio astronomical images

## TL;DR

This paper introduces a highly efficient multi-scale deconvolution algorithm for radio astronomical images, significantly faster than existing methods, with extensions for multi-frequency mode and automated masking, improving image quality and processing speed.

## Contribution

The paper presents a novel multi-scale deconvolution algorithm that is faster and more flexible than CASA's, including multi-frequency mode and automated masking, with a comparison to a convex optimisation approach.

## Key findings

- Minor loop of the new algorithm is over an order of magnitude faster in single-frequency mode.
- In multi-frequency mode, the algorithm is 2-3 orders of magnitude faster than CASA MSMFS.
- Convex optimisation produces better models on simple data but has stability issues on complex data.

## Abstract

We describe a new multi-scale deconvolution algorithm that can also be used in multi-frequency mode. The algorithm only affects the minor clean loop. In single-frequency mode, the minor loop of our improved multi-scale algorithm is over an order of magnitude faster than the CASA multi-scale algorithm, and produces results of similar quality. For multi-frequency deconvolution, a technique named joined-channel cleaning is used. In this mode, the minor loop of our algorithm is 2-3 orders of magnitude faster than CASA MSMFS. We extend the multi-scale mode with automated scale-dependent masking, which allows structures to be cleaned below the noise. We describe a new scale-bias function for use in multi-scale cleaning. We test a second deconvolution method that is a variant of the MORESANE deconvolution technique, and uses a convex optimisation technique with isotropic undecimated wavelets as dictionary. On simple, well calibrated data the convex optimisation algorithm produces visually more representative models. On complex or imperfect data, the convex optimisation algorithm has stability issues.