Skip to content
DraconiticResearch
Search Sources API
203,234 studies31 sourcesUpdated 9 min ago
Peer-reviewedCC BY 4.0The Astrophysical Journal

Pixelated Posterior Sampling of Point-spread Functions in Astronomical Images

  • Connor Stone1,2,3,4
  • Ronan Legin1,2,3
  • Alexandre Adam1,2,3
  • Esmeralda Whitammer2,3,5
  • Gabriel Missael Barco1,2,3
  • Laurence Perreault-Levasseur1,2,3,6,7,8
  • Yashar Hezaveh1,2,3,6,7,8
8 affiliations
  1. 1Ciela—Montreal Institute for Astrophysical Data Analysis and Machine Learning
  2. 2Université de Montréal
  3. 3Mila - Quebec Artificial Intelligence Institute
  4. 4University of Toronto
  5. 5University of Edinburgh
  6. 6Flatiron Institute
  7. 7Trottier Space Institute
  8. 8Perimeter Institute
Open at sourcePDFDOICite

Abstract

We introduce a novel framework for upsampled point-spread function (PSF) modeling using pixel-level Bayesian inference. Accurate PSF characterization is critical for precision measurements in many fields, including weak lensing, astrometry, and photometry. Our method defines the posterior distribution of the pixelized PSF model through the combination of an analytic Gaussian likelihood and a highly expressive generative diffusion model prior, trained on a library of Hubble Space Telescope effective PSF (ePSF) templates. Compared to traditional methods (parametric Moffat, ePSF template-based, and regularized likelihood), we demonstrate that our PSF models achieve orders-of-magnitude-higher likelihood and residuals consistent with noise, all while remaining visually realistic. Further, the method applies even for faint and heavily masked point sources, merely producing a broader posterior. By recovering a realistic, pixel-level posterior distribution, our technique enables the first meaningful propagation of detailed PSF morphological uncertainty in downstream analysis. An implementation of our posterior sampling procedure is available on GitHub.

Related

  1. Aperture Photometric Accuracy of Point-spread-function-deconvolved Astronomical Images
    Shen Zhang, Lei Wang, Yachong Diao +11
    arXivPreprintCC BY 4.0Full textInstrumentation & methodsarXiv
  2. POSTELLAR: Posterior Stellar Spectrum Sampling - An Alternative to Approximate Stellar Spectra for Exoplanetary Analysis
    Dhvani Doshi, Nicolas B. Cowan, Yashar Hezaveh +2
    arXivPreprintarXiv licenceFull textInstrumentation & methodsPlanetary science & exoplanetsarXiv
  3. The Impact of Non-Gaussian Line Spread Functions on Stellar Kinematic Recovery: Consequences for Dynamical Models
    David A. Simon
    arXivPreprintCC BY 4.0Full textGalaxiesarXiv
  4. A Missing Tool for Calculating the Auto/Cross-correlation Functions under Nonuniform Sampling Observations
    Chen-Ran 宸然 Hu 胡, Yong-Feng 永锋 Huang 黄, Jin-Jun 金军 Geng 耿 +10
    The Astrophysical Journal Supplement SeriesPeer-reviewedCC BY 4.0Full textSolar & stellar physicsCrossref
  5. Active focal plane phase mask coronagraphy with a discrete pixelated device: I. Study of the theoretical performance trade space
    Liurong Lin, Axel Potier, Ruben Tandon +2
    arXivPreprintarXiv licenceFull textInstrumentation & methodsPlanetary science & exoplanetsarXiv

Cite

BibTeXRIS
BibTeX
@article{stone2026pixelated,
  title = {{Pixelated Posterior Sampling of Point-spread Functions in Astronomical Images}},
  author = {Connor Stone and Ronan Legin and Alexandre Adam and Esmeralda Whitammer and Gabriel Missael Barco and Laurence Perreault-Levasseur and Yashar Hezaveh},
  journal = {The Astrophysical Journal},
  year = {2026},
  month = oct,
  doi = {10.3847/1538-4357/aea97f},
  url = {https://iopscience.iop.org/article/10.3847/1538-4357/aea97f},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - Pixelated Posterior Sampling of Point-spread Functions in Astronomical Images
AU  - Connor Stone
AU  - Ronan Legin
AU  - Alexandre Adam
AU  - Esmeralda Whitammer
AU  - Gabriel Missael Barco
AU  - Laurence Perreault-Levasseur
AU  - Yashar Hezaveh
PY  - 2026
DA  - 2026/10/29
JO  - The Astrophysical Journal
DO  - 10.3847/1538-4357/aea97f
UR  - https://iopscience.iop.org/article/10.3847/1538-4357/aea97f
AB  - We introduce a novel framework for upsampled point-spread function (PSF) modeling using pixel-level Bayesian inference. Accurate PSF characterization is critical for precision measurements in many fields, including weak lensing, astrometry, and photometry. Our method defines the posterior distribution of the pixelized PSF model through the combination of an analytic Gaussian likelihood and a highly expressive generative diffusion model prior, trained on a library of Hubble Space Telescope effective PSF (ePSF) templates. Compared to traditional methods (parametric Moffat, ePSF template-based, and regularized likelihood), we demonstrate that our PSF models achieve orders-of-magnitude-higher likelihood and residuals consistent with noise, all while remaining visually realistic. Further, the method applies even for faint and heavily masked point sources, merely producing a broader posterior. By recovering a realistic, pixel-level posterior distribution, our technique enables the first meaningful propagation of detailed PSF morphological uncertainty in downstream analysis. An implementation of our posterior sampling procedure is available on GitHub.
SN  - 1538-4357
SN  - 0004-637X
ER  -