Direct model fitting to combine dithered ACS images

TL;DR

This paper proposes a direct model fitting method for combining dithered images from undersampled CCD cameras, improving the accuracy of gravitational lensing analyses by avoiding artifacts from traditional combination techniques.

Contribution

It introduces a novel direct model fitting approach that incorporates dithering and lensing effects without creating intermediate combined images, reducing systematic errors.

Findings

Method produces high-quality combined images in tests.

Incorporates gravitational lensing and PSF effects naturally.

Reduces artifacts compared to traditional methods.

Abstract

The information lost in images of undersampled CCD cameras can be recovered with the technique of `dithering'. A number of subexposures is taken with sub-pixel shifts in order to record structures on scales smaller than a pixel. The standard method to combine such exposures, `Drizzle', averages after reversing the displacements, including rotations and distortions. More sophisticated methods are available to produce, e.g., Nyquist sampled representations of band-limited inputs. While the combined images produced by these methods can be of high quality, their use as input for forward-modelling techniques in gravitational lensing is still not optimal, because the residual artefacts still affect the modelling results in unpredictable ways. In this paper we argue for an overall modelling approach that takes into account the dithering and the lensing without the intermediate product of a combined image. As one building block we introduce an alternative approach to combine dithered images by direct model fitting with a least-squares approach including a regularization constraint. We present tests with simulated and real data that show the quality of the results. The additional effects of gravitational lensing and the convolution with an instrumental point spread function can be included in a natural way, avoiding the possible systematic errors of previous procedures.