[Paper Review] MegaPipe: the MegaCam image stacking pipeline

Stephen Gwyn|ArXiv.org|Apr 16, 2009

Astronomical Observations and Instrumentation3 citations

TL;DR

MegaPipe is an automated image processing pipeline that combines multiple MegaCam images from CFHT into single, astrometrically and photometrically calibrated stacks. It uses SExtractor for source detection, SDSS DR7 for photometric calibration, and SWarp for coaddition, producing 1.7 GB FITS stacks with 0.025'' astrometric accuracy and 0.02–0.03 mag photometric accuracy, enabling efficient access to 1500 square degrees of processed data.

ABSTRACT

This paper describes the MegaPipe image processing pipeline at the Canadian Astronomical Data Centre (CADC). The pipeline takes multiple images from the MegaCam mosaic camera on CFHT and combines them into a single output image. MegaPipe takes as input detrended MegaCam images and does a careful astrometric and photometric calibration on them. The calibrated images are then resampled and combined into image stacks. MegaPipe is run on PI data by request, data from large surveys (the CFHT Legacy Survey and the Next Generation Virgo Survey) and all non-proprietary MegaCam data in the CFHT archive. The stacked images and catalogs derived from these images are available through the CADC website. Currently, 1500 square degrees have been processed.

Motivation & Objective

• To reduce the barrier to using archival MegaCam data, which are large, distorted, and require significant processing effort.
• To automate the calibration and coaddition of MegaCam images into single, high-quality, distortion-free stacks.
• To ensure consistent astrometric and photometric accuracy across large sky areas using external catalogs and robust statistical methods.
• To make processed MegaCam data and derived catalogs publicly accessible through the CADC website and APIs.

Proposed method

• Input images are visually inspected and grouped using a friends-of-friends algorithm with a 0.1° separation threshold.
• SExtractor detects sources in each image, which are cleaned of cosmic rays and extended objects to isolate stars and compact galaxies.
• Astrometric solutions are derived using external catalogs (USNO A2 or SDSS DR7), with global radial distortion modeled via second- and fourth-order polynomial terms.
• Photometric calibration uses SDSS DR7 as a reference, applying median zero-point offsets between instrumental and SDSS magnitudes for 17 < mag < 20 stars.
• Images are coadded using SWarp into 20,000×20,000 pixel FITS files with a photometric zero-point of 30.000 AB mag and inverse-variance weight maps.
• Catalogs are generated via SExtractor on each stack, and calibration accuracy is validated using cross-matching and residual analysis.

Experimental results

Research questions

• RQ1 Can automated image stacking significantly reduce the effort required to use archival MegaCam data?
• RQ2 What level of astrometric and photometric accuracy can be achieved when calibrating MegaCam images using SDSS DR7 as a reference?
• RQ3 How effectively can a global distortion model reduce the number of parameters needed for accurate astrometric calibration across a 36-CCD mosaic?
• RQ4 What is the limiting magnitude and surface brightness sensitivity of the stacked images after coaddition?
• RQ5 How repeatable and reliable are the astrometric and photometric solutions across overlapping fields and different observing conditions?

Key findings

• The internal astrometric uncertainty of the calibrated stacks is 0.025 arcseconds, indicating high positional precision.
• The astrometric repeatability between overlapping catalogs is 0.06 arcseconds, demonstrating consistency across independent processing.
• External astrometric accuracy, after accounting for reference catalog uncertainties, is 0.2 arcseconds, confirming reliable sky alignment.
• The photometric accuracy is 0.02–0.03 magnitudes, with a systematic offset of 0.015 magnitudes between independent pointings.
• The limiting magnitude is determined via three methods: peak number counts, 5σ detection threshold (faintest object with magnitude error ≤0.198 mag), and fake-source injection, yielding robust depth estimates.
• The pipeline successfully processed 1500 square degrees of public MegaCam data, with outputs distributed via CADC’s website and image cutout service.