{
  "abstract": "Aims. Dusty star-forming galaxies (DSFGs) dominate the far-infrared (FIR) and sub-millimetre (sub-mm) number counts, but singledish surveys at these wavelengths suffer from poor angular resolution, making identifications of multi-wavelength counterparts difficult. Prior driven deblending techniques require extensive fine-tuning and struggle to process large fields. This work aims to develop a fast and reliable deep-learning-based deconvolution and denoising super-resolution (SR) technique. Methods. We employed a transformer neural network to improve the resolution of the Herschel/SPIRE 500 µm observations by a factor of 4.5, with input comprised of Spitzer/MIPS 24µm and Herschel/SPIRE 250, 350, 500 µm images. The network was trained on simulations from SIDES and SHARK. To mimic realistic observations, we injected instrumental noise into the input simulated images, while keeping the target images noise-free to enhance the de-noising capabilities of our method. We evaluated the performance of our method on simulated test sets and real JCMT/SCUBA-2 450 µm observations in the COSMOS field that have a superior resolution compared to Herschel. Results. Our SR method achieves an inference time of ~1 s/deg2 on consumer-grade GPUs, which is much faster than traditional deblending techniques. Using the simulation test sets, we show that fluxes of the extracted sources from the super-resolved image are accurate to within 5% for sources with an intrinsic flux ≳8 mJy, which is a substantial improvement compared to blind extraction on the native images. Astrometric error is low, at ≲1″ compared to the 12″ pixel scale. In terms of reliability and completeness, ≳90% of the extracted sources brighter than ~3 mJy are reliable and more than 90% of the input sources with intrinsic fluxes ≳5 mJy are recovered. When applied to the real 500 µm observations, the fluxes of the extracted sources from the super-resolved map agree well with the SCUBA-2 measured fluxes (after converting the 450 µm fluxes to 500 µm using a correction factor of 0.84) for sources above ~10 mJy. Our 500 µm number counts are also consistent with previous SCUBA-2 measurements. Thanks to its speed, our technique enables SR over hundreds of deg2 without the need for fine-tuning, facilitating statistical analysis of DSFGs.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": null,
  "author_total": 12,
  "authors": [
    {
      "name": "Dennis Koopmans"
    },
    {
      "name": "Lingyu Wang"
    },
    {
      "name": "Berta Margalef-Bentabol"
    },
    {
      "name": "Antonio La Marca"
    },
    {
      "name": "Matthieu Bethermin"
    },
    {
      "name": "Laura Bisigello"
    },
    {
      "name": "Zhen-Kai Gao"
    },
    {
      "name": "Claudia del P. Lagos"
    },
    {
      "name": "Lynge Lauritsen"
    },
    {
      "name": "Stephen Serjeant"
    },
    {
      "name": "F F S van der Tak"
    },
    {
      "name": "Wei-Hao Wang"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/34bsbj1qwetjp.bib",
  "cited_by": 0,
  "doi": "10.1051/0004-6361/202555943",
  "fulltext_url": null,
  "id": "34bsbj1qwetjp",
  "issns": [
    "0004-6361",
    "1432-0746"
  ],
  "landing_url": "https://hal.science/hal-05778672v1",
  "language": "en",
  "licence": "CC-BY-4.0",
  "licence_label": "CC BY 4.0",
  "licence_source": "article",
  "licence_url": "https://creativecommons.org/licenses/by/4.0/",
  "metadata_licence": null,
  "native_subjects": [
    [
      "hal",
      "0.sdu"
    ],
    [
      "hal",
      "1.sdu.astr"
    ],
    [
      "hal",
      "2.sdu.astr.ga"
    ],
    [
      "openalex",
      "1707|Advanced Image Processing Techniques"
    ]
  ],
  "openalex": "W7167608217",
  "pdf_url": "https://hal.science/hal-05778672/document",
  "pmcid": null,
  "pmid": null,
  "published": "2026-10-05",
  "publisher": "EDP Sciences",
  "ris": "https://research.draconitic.com/paper/34bsbj1qwetjp.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:17:47Z",
      "native_id": "hal-05778672",
      "source": "hal",
      "url": "https://hal.science/hal-05778672v1"
    }
  ],
  "subjects": [
    {
      "field": "astronomy",
      "origin": "hal:2.sdu.astr.ga",
      "subfield": "galaxies"
    },
    {
      "field": "earth",
      "origin": "hal:0.sdu",
      "subfield": "general"
    }
  ],
  "title": "Super-resolving Herschel: A deep-learning-based deconvolution and denoising technique: I. First application at 500 µm in the COSMOS field",
  "title_text": "Super-resolving Herschel: A deep-learning-based deconvolution and denoising technique: I. First application at 500 µm in the COSMOS field",
  "type": "peer-reviewed",
  "updated_at": "2026-10-30T12:17:47Z",
  "url": "https://research.draconitic.com/paper/34bsbj1qwetjp",
  "venue": "Astronomy & Astrophysics - A&A",
  "versions": []
}