Point spread function wavefront recovery from in-focus stellar observations

summary

The gist

Recovering the wavefront error (WFE) field of an optical system from intensity in-focus observations is a challenging inverse problem with broad implications for telescope point spread function (PSF)…

In short

The paper addresses recovering wavefront errors from blurry images using a model called WaveDiff. It introduces a new optimization method that connects the parametric and non-parametric parts of WaveDiff through wavefront feature projection. This significantly reduces the error in estimating the wavefront error field, improving accuracy by nearly an order of magnitude.

Key concepts

Point Spread Function (PSF)
The PSF describes how light from a distant object is blurred by the telescope's optical system. It tells scientists exactly how much the telescope distorts and smears an image, which is crucial for accurately measuring cosmological data.
WaveDiff Model
This is a semi-parametric approach used to model the PSF in terms of wavefront space. It combines a simple, predictable parametric component with a flexible non-parametric component to better represent the complex optical distortions.
Wavefront Feature Projection
This is a new technique that links the information from the flexible non-parametric part of WaveDiff to its fixed parametric part. This projection helps exploit spatial variations in optical aberrations, leading to a much more accurate reconstruction of the wavefront error field.

Terminology used across episodes

This episode discusses

The paper

Point spread function wavefront recovery from in-focus stellar observations · Read on arXiv

Ezequiel Centofanti, Samuel Farrens, Jean-Luc Starck, Tobías I. Liaudat

AIM · CEA CNRS Université Paris-Saclay University of Paris Institute of Computer Science and Astrophysics Foundation for Research and Technology Hellas

Recovering the wavefront error (WFE) field of an optical system from intensity in-focus observations is a challenging inverse problem with broad implications for telescope point spread function (PSF) modelling. Accurate WFE recovery enables both precise PSF modelling and direct insight into the state of the telescope optics, facilitating the detection of potential malfunctions. Recently, non-parametric PSF models have shown promising performance in modelling complex optical systems in space-based telescopes. WaveDiff is a semi-parametric PSF model that represents the PSF in wavefront space by combining parametric and learnable features with a differentiable forward optical model. This parameterisation enables phase retrieval from in-focus observations by exploiting the spatial variation of the PSF across the field of view (FOV). The original version of WaveDiff achieves outstanding PSF recovery results in pixel space; however, the recovered WFE is far from the ground truth, with a relative error of around 30 %. In this paper, we present a new optimisation scenario that bridges WaveDiff's parametric and non-parametric components through wavefront feature projection, yielding a substantial improvement in WFE recovery and making WaveDiff the first demonstrated method to combine wide-field WFE recovery, in-focus-only polychromatic observations, and non-parametric wavefront features in a single framework. We show that incorporating wavefront projections and increasing the number of optimisation cycles enables WaveDiff to recover the WFE with an error of approximately 3 % using only noisy, undersampled, in-focus observations. This represents a tenfold improvement over the original model while further reducing the pixel-space error. The code to reproduce the results of this article is publicly available at https://github.com/tobias-liaudat/wf-psf/tree/v1.4.0

DOI: 10.1051/0004-6361/202558730

Transcript

Introduction to the show: ident: Astrophysics. The day in astrophysics: observations, models and what the data forces.

Vera: We're looking at the paper: "Point spread function wavefront recovery from in-focus stellar observations".

Jocelyn: Here's the gist of the paper. Recovering the wavefront error (WFE) field of an optical system from intensity in-focus observations is a challenging inverse problem with broad implications for telescope point spread function (PSF) modelling.

Vera: This paper presents a new optimization scenario for the WaveDiff semi-parametric PSF model that bridges its parametric and non-parametric components through wavefront feature projection, yielding a substantial improvement in WFE recovery and making WaveDiff the first demonstrated method to combine wide-field WFE recovery, in-focus-only polychromatic observations, and non-parametric wavefront features in a single framework.

Jocelyn: The PSF Modelling Problem The PSF characterises the effects that the optical system introduces to the image of an observed object.

Vera: The instrumental response, commonly referred to as the point spread function (PSF), describes how incident light is blurred as it passes through the optical system of the telescope.

Paper summary: Vera: So, as we talked about, the core idea here is recovering the wavefront error field using intensity in-focus observations, which is a tough inverse problem because you are trying to figure out something invisible based on what you see. The paper introduces WaveDiff as a semi-parametric PSF model that represents this PSF in wavefront space and fits it only with in-focus stars.

Jocelyn: They break down the wavefront error representation into two parts, a parametric term and a non-parametric term, which they then link together using some new feature projections. This linkage is what they say allows them to exploit the spatial diversity of optical aberrations better than before.

Subrahmanyan: The mechanism they use to transfer information from the non-parametric part back to the parametric part involves an iterative projection algorithm that handles what they call Zernike mixing introduced by the obscuration mask. This suggests they are dealing with some kind of non-orthogonality errors in their model setup.

Vera: That process allows them to include a spatially varying obscuration mask, which is something that’s important for real telescopes because the optical path changes depending on where the light comes from across the field of view.

Jocelyn: They also note that modeling the PSF in wavefront space enables them to model how the spectral variation of the PSF changes across an entire field of view, which they say is something no other non-parametric model is currently capable of doing.

Subrahmanyan: That spectral variation due to diffraction across the whole field really connects this work to observations like weak lensing studies, because those studies need a good PSF model even when applying it to galaxies with different spectral energy distributions.

Vera: The paper states that this telescope wavefront error estimation doesn't just build a precise PSF model; it also lets you diagnose potential issues in the optical system itself. They cite something like the James Webb Space Telescope using wavefront sensing to detect things like micrometeorite impacts on its mirrors.

Jocelyn: So, they show that by modeling the PSF this way, we get two major things: a very precise model for analysis and a diagnostic tool for checking if the telescope hardware is behaving correctly.

Subrahmanyan: It’s about using in-focus stellar observations to constrain these complex models, which is valuable because stars are used to constrain the PSF model, but you have to account for those variations caused by refractive elements and dielectric coatings as well.

Conclusion: Vera: So, looking at the whole "Point spread function wavefront recovery from in-focus stellar observations" paper, the main thing is that they developed a new optimization scenario for WaveDiff. This scenario bridges the parametric and non-parametric components of the model through wavefront feature projection.

Jocelyn: The authors claim this new setup yields a substantial reduction in the wavefront error error, dropping it from about thirty percent down to three point four percent, which is nearly an order of magnitude better than their previous work.

Subrahmanyan: For me, the big picture is that this provides a much more accurate way to characterize the optical imperfections of these instruments when we rely on in-focus stellar data. It gives us a tighter constraint on how the telescope optics are actually behaving across its field of view.

Vera: And it makes WaveDiff one of the first methods demonstrated to successfully combine wide-field wavefront error recovery, in-focus only observations, and non-parametric wavefront features all in a single framework.

Jocelyn: It’s an important step because it shows that you can get these different types of information—the parametric and the non-parametric—to work together effectively for this kind of problem.

Subrahmanyan: Ultimately, this means we have a much better tool to diagnose potential malfunctions in large optical systems by using simple stellar observations, which is a significant practical advance for future telescope operations.

More episodes

← Home