DRUID: Source Detection and Deblending in Astronomical Images with Persistent Homology

TL;DR

DRUID is a novel source detection method using persistent homology that effectively segments complex and nested astronomical sources in optical and radio images, improving deblending in high-density surveys.

Contribution

The paper introduces DRUID, a new source finder leveraging persistent homology for improved detection and deblending of complex, nested sources in astronomical images.

Findings

Successfully segments complex radio galaxy morphologies

Effectively deblends sources in high-density images

Produces detailed source catalogues for large surveys

Abstract

Source detection is a vital part of any astronomical survey analysis pipeline. In addition, a versatile source finder that can recover and handle sources of all morphological types is becoming more important as surveys get bigger and achieve a higher resolution than ever before. Here we present Detector of astRonomical soUrces in optIcal and raDio images (DRUID), a source finder that utilises persistent homology to detect and deblend sources. This method enables us to effectively and uniquely segment structures within morphologically complex sources and deal with high source density images. We test DRUID on the complex morphologies of 3CR radio loud active galactic nuclei, where we demonstrate its ability to usefully segment the main structures in the sources. We also demonstrate the level of structure DRUID segments within well resolved galaxies in the optical. Finally, we present two source catalogues on the LoTSS Deep field observation of the Lockman Hole and an example tile from the KiDS r-band survey. We conclude that DRUID's method of utilising persistent homology provides a new way to detect and deblend highly nested sources.