Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A review on fundamental bounds and estimators for photometry and astrometry of celestial point sources using array detectors, from first principles

Sebastián Espinosa, Rene A. Mendez, Jorge F. Silva, Marcos Orchard

TL;DR

This paper surveys the theoretical and practical limits of astrometry and photometry for celestial point sources using array detectors under Poisson statistics. It develops a unified observational model, derives CRLB-based bounds for one- and multi-parameter estimation (including moving sources), and systematically compares estimators such as LS, WLS, AWLS, and ML, highlighting where ML and adaptive approaches achieve or approach the bound. The work also discusses Bayesian extensions (BCRLB), joint estimation of flux and background, and real-data validations, demonstrating the practical impact on modern surveys (e.g., Gaia, LSST, TESS). The overarching message is that estimator design guided by Fisher information leads to more robust, near-optimal photometric and astrometric pipelines, with clear guidance for handling complex PSFs, sampling, and dynamic observing conditions in future large-scale surveys.

Abstract

Precise astrometric and photometric measurements of celestial point sources are fundamental to modern astronomy. These measurements, used to determine object positions, motions, and fluxes, are based on observational models that have evolved from empirical centroiding rules to rigorous probabilistic formulations at the pixel level. This review summarizes key contributions that formalized this transition and analyzes seminal works addressing both the theoretical limits and the empirical performance of estimators. Central to these developments is the derivation of fundamental bounds, such as the Cramér-Rao Lower Bound (CRLB), and the assessment of widely used estimators, including Maximum Likelihood (ML), Least Squares (LS), and Weighted Least Squares (WLS). These studies show that, while the CRLB sets a theoretical benchmark, practical estimators achieve it only under specific signal-to-noise ratio (SNR) regimes, with notable discrepancies in high-SNR conditions. Moreover, recent results demonstrate that jointly estimating source flux and background significantly improves photometric precision compared to sequential approaches. Looking ahead, the increasing complexity of astronomical surveys, driven by massive data volumes, dynamic observational conditions, and the integration of machine learning, poses new challenges to reliable inference. In this context, tools from statistical theory, including performance bounds and theoretically grounded estimators, remain critical to guide algorithm design and ensure robust astrometric and photometric pipelines.

A review on fundamental bounds and estimators for photometry and astrometry of celestial point sources using array detectors, from first principles

TL;DR

This paper surveys the theoretical and practical limits of astrometry and photometry for celestial point sources using array detectors under Poisson statistics. It develops a unified observational model, derives CRLB-based bounds for one- and multi-parameter estimation (including moving sources), and systematically compares estimators such as LS, WLS, AWLS, and ML, highlighting where ML and adaptive approaches achieve or approach the bound. The work also discusses Bayesian extensions (BCRLB), joint estimation of flux and background, and real-data validations, demonstrating the practical impact on modern surveys (e.g., Gaia, LSST, TESS). The overarching message is that estimator design guided by Fisher information leads to more robust, near-optimal photometric and astrometric pipelines, with clear guidance for handling complex PSFs, sampling, and dynamic observing conditions in future large-scale surveys.

Abstract

Precise astrometric and photometric measurements of celestial point sources are fundamental to modern astronomy. These measurements, used to determine object positions, motions, and fluxes, are based on observational models that have evolved from empirical centroiding rules to rigorous probabilistic formulations at the pixel level. This review summarizes key contributions that formalized this transition and analyzes seminal works addressing both the theoretical limits and the empirical performance of estimators. Central to these developments is the derivation of fundamental bounds, such as the Cramér-Rao Lower Bound (CRLB), and the assessment of widely used estimators, including Maximum Likelihood (ML), Least Squares (LS), and Weighted Least Squares (WLS). These studies show that, while the CRLB sets a theoretical benchmark, practical estimators achieve it only under specific signal-to-noise ratio (SNR) regimes, with notable discrepancies in high-SNR conditions. Moreover, recent results demonstrate that jointly estimating source flux and background significantly improves photometric precision compared to sequential approaches. Looking ahead, the increasing complexity of astronomical surveys, driven by massive data volumes, dynamic observational conditions, and the integration of machine learning, poses new challenges to reliable inference. In this context, tools from statistical theory, including performance bounds and theoretically grounded estimators, remain critical to guide algorithm design and ensure robust astrometric and photometric pipelines.

Paper Structure

This paper contains 37 sections, 8 theorems, 42 equations, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Modeling evolution in astrometry and photometry
  3. Foundations of the observational model
  4. Mathematical formulation
  5. Source flux and noise contributions
  6. Point spread function (PSF) and sampling
  7. The inference task
  8. Practical implications for astrometric and photometric estimation
  9. Practical aspects of PSF modeling and recent advances
  10. Analytic representations
  11. Empirical pixel-integrated reconstructions
  12. Data-driven representations
  13. Current perspective
  14. Fundamental bounds in astrometry and photometry
  15. Early contributions in precision limits
  16. ...and 22 more sections

Key Result

Theorem 1

rao1945cramer1946 Let $\mathcal{N}$ a finite set of integers with $|\mathcal{N}|=n$, $\mathbf{I}=\{I_{i}\}_{i \in \mathcal{N}}$ be a collection of observations, whose likelihood function $L(\cdot; \boldsymbol{\theta})$ is induced by a parameter vector $\boldsymbol{\theta} = (\theta_{1},\dots,\theta_ (i) Then, any unbiased estimator $\hat{\boldsymbol{\theta}}$ of $\boldsymbol{\theta}$, given by a r

Figures (4)

  • Figure 1: Illustration of a point source emitting flux onto a CCD array. The resulting intensity distribution on the detector follows the PSF, whose centroid corresponds to the position of the source.
  • Figure 2: Pixel-integrated model of the observed intensity. The expected photon count at pixel $i$ is given by $\lambda_i(x_c, F) = F \cdot g_i(x_c) + B$, where $g_i(x_c)$ integrates the PSF over pixel $i$, and $B$ accounts for the background.
  • Figure 3: Square root of the CRLB for centroid estimation as a function of the pixel size $\Delta x$. Parameters correspond to a typical ground-based observation: total source flux $F=1000\ e^-$, background level $B=10\ e^-$ per pixel, Gaussian PSF with standard deviation $\sigma = 0.8$ arcsec, centroid $x_c=1.12$ pixels, and a well-sampled window of $n=201$ pixels. The CRLB is expressed in milliarcseconds (mas). The curve shows the characteristic U-shape, with an optimum region in $\Delta x$ that minimizes the astrometric uncertainty.
  • Figure 4: Illustration of a dynamic PSF for a moving source. The PSF is elongated along the trajectory direction due to the linear motion of the source during the exposure. The color gradient indicates intensity, with warmer colors representing higher flux levels. This schematic is inspired by bouquillon2017characterizing but adapted for illustrative purposes.

Theorems & Definitions (8)

  • Theorem 1
  • Lemma 2
  • Theorem 3
  • Theorem 4
  • Theorem 5
  • Theorem 6
  • Theorem 7
  • Theorem 8