2004 Interferometry Imaging Beauty Contest


WG Home | 2001 | 2002 | 2003 | 2004 | Notes
Data Exchange Format | CFITSIO | IDL Utilities
mfit | OYSTER | ASPRO | VisCalc
Contest Data Set 1: data1.oifits
Contest Data Set 2: data2.oifits
The IAU Working Group on Optical/IR Interferometry is organizing an Imaging Beauty Contest whose results will be presented at the SPIE Conference on New Frontiers in Stellar Interferometry. The interferometry conference is part of the Conference on Astronomical Telescopes and Instrumentation 2004, 21-25 June 2004, Glasgow, Scotland, United Kingdom.
The contest is being chaired by Peter Lawson (Jet Propulsion Laboratory) with participation from the following:

Organizers

William Cotton
NRAO
Coordinator for Test Data
Christian Hummel
European Southern Observatory
Test Data generated through Oyster

Contributors

Marc Kuchner
Princeton University
Test Data: Models
Gilles Duvert
JMMC, Grenoble
Test Data: ASPRO
Paulo Garcia
Centro de Astrofisica da Universidade do Porto
Test Data: Model Images

Participants

Name
Institution
Software
Status
John Young
Hrobjartur Thorsteinsson
University of Cambridge
BSMEM
Tests Complete
John Monnier
University of Michigan
IDL
Tests Complete
Eric Thiebault
Observatoire de Lyon
Regularized Image
Reconstruction

Serge Meimon
Laurent Mugnier
Guy le Besnerais
ONERA
Weak Phase Image
Reconstruction (IDL)

Motivation for the Contest

The IAU Working Group on Optical/IR Interferometry seeks to encourage international collaboration in the field of long-baseline stellar interferometry. At the 2001 Meeting of the Working Group, the state-of-the-art in imaging algorithms was discussed. The differences between radio and optical imaging were reviewed by David Buscher (see his presentation) with an emphasis on the needs of optical interferometry and the suggestion that existing algorithms should be compared to better understand observing strategies and the limitations of optical imaging.
Since the 2001 Meeting, the Data Exchange Standard for Optical/IR Interferometry was established, principally through the efforts of Tom Pauls (NRL) and John Young (Cambridge University). The Data Exchange Standard is now supported at NPOI, COAST, CHARA, IOTA, the Michelson Science Center (Keck and PTI), and the VLTI. IDL Utilities, developed by John Monnier (University of Michigan), are now also available.
In 2003 the development of the Exchange Standard has begun to facilitate collaborative science between interferometry groups and now also facilitates a comparative assessment of imaging algorithms.
There is willingness amongst prospective participants (noted above), to evaluate imaging algorithms through a formal comparison and to encourage the more widespread use of the Exchange Standard. The goals and metrics of evaluation for the contest, however, are yet to be decided and will be established by discussions amongst the participants.
The participants will, through mutual agreement, set the ground-rules for the competition in October and November 2003. The contest Chair will by November 24th submit an abstract to the 2004 Interferometry Conference with himself as lead author, and include all participants as co-authors.

Optical Long-baseline Interferometry Data Sets

The data sets for the contest should be relevant to concerns particular to long-baseline stellar interferometry. There should be a limited number of data sets that each test something quite specific.
In some statistical sense the data sets should faithfully represent data from plausible long-baseline stellar interferometers. The following characteristics may be considered typical:
  • All currently planned arrays will sample significantly fewer data points than Keck aperture masking. The contest data should therefore not have as extensive sampling as Keck aperture-masking data.
  • The contest data should probably have about N(N-1)/2 u-v points per hour of observing and fewer bispectrum points. This would be consistent with array of 3 or 4 apertures, reconfigured one or more times.
  • The data should be mostly in the low SNR regime.
  • The observables will be based on power spectra and bispectra.
  • The relationship between u-v coverage and bispectrum will not be as straightfordard as in the radio. VLBI algorithms/software should not be not well suited to reduce the optical long-baseline data used in the contest.
  • The test data should probably represent a source with a complicated symmetry so that measurements of closure phases are essential for image reconstruction. Parametric imaging will not recover all of the source structure in the abscence of a priori knowledge.

What to Test For?

There are a lot of subjects bundled up in the above notes. It might be useful to try to separate some of the issues and not confuse several topics. It would be very useful to identify the most pressing concerns and tailor the example data to best represent a known problem that we can then test for. I would appreciate your input on this. The subjects that come to mind based on the above notes are as follows:
  • Squared-visibilities but missing closure phases on X% of data?
  • Closure phases but missing squared-visibilities on Y% of data?
Other possible topics:
  • Problems with non-zero closure phase errors.
  • Response to point source in an extended structure.
  • Cross-talk in bispectra.

Proposed Contest Rules (To Be Discussed by Participants)

The likely ground-rules, to be discussed, are as follows:

Imaging Data products

  • Calibrated imaging data supplied in the Data Exchange Format.
  • Common data sets supplied to all participants.
  • Visibilities-squared, closure phases, and closure amplitudes provided.
  • Baseline phases not provided.
The data sets may have a high signal-to-noise ratio, but include structure with a broad dynamic range of intensities. The data will be simulated at a single wavelength. The data may also include realistic noise, calibration errors, and missing closure phases or squared visibilities.

Source Morphology

  • Double star provided as test object - revealed to contestants
  • Mystery objects may include
    • Limb darkened star with one or more spots
    • Compact source with extended envelope
    • Other more exotic objects

Array Configuration

  • 3 to 4 telescope Y-shaped array similar to NPOI, COAST, or CHARA.
  • Array re-configurations, baselines, and uv coverage to be decided.
The beauty contest will proceed in several rounds starting with a simple model such as an assymetric double star with good SNR and no serious complications in the data. This will work out the bumps in the data distribution process and locate any differences in conventions. Once the results for this are satisfactory, a subsequent round can add complications.
The model for the first round could be provided with the data as the purpose of this round is to ensure that everybody can read the Exchange format and that all the sign and coordinate system conventions etc. are consistent. In the subsequent round(s), the model would not be revealed until all groups had submitted their best results.

A fits image of a YSO jet model (not in the OIDATA standard) from Paulo Garcia can be found in
M82_beaty_contest.fits.gz
His description if the image is: I can generate a fits file of an image, it is standard fits but it is not the interferometry standard. The angular size of the image depends on the distance of the object. The fits image I'm attaching is the synthetic image in an emission line of a jet in a YSO (there is no continnum emission from the jet). The scale is 0.09 AU/pix, for a YSO at Taurus (140pc) the pixel scale is 0.643 milliarcsec/pix. As you will see the jet is *huge*, however most of it is very faint, too faint for an optical interferometer. If you are optimistic and only see the flux up to around 1% of the peak, then the jet becomes very small. This is what we used in our paper (attached). As it is a jet image, the star is missing. You could expect a disk, but some calculations for disks from low mass YSOs show that they have negligible emission at these spatial scales and at the J band considered.

Christian Hummel has produced two files (attached) in the OI-FITS format of simulated binary data with a simulated 6-station array, 15 baselines, and 10 triples.
  • Test Data Set 1: BSC1948.fits. Can be modelled by a binary with sep. rho=21.2 mas and pa=341.6 deg, flux ratio 5.75, component diameters of 0.6 mas. This information is being made available so you can check if you are able to fit the correct model parameters.
  • Test Data Set 2: FKV1137.fits. Also a binary, but more complicated.
Contestants should be able to produce an image with the correct model parameters by 5 March 2004.

Christian has now made available the contest data. There are two data sets, as follows. Please note that the target names cited in the oi-fits files (Mizar and Altair) bear no relation to the data itself.
As noted in the schedule below, please provide your final images to the contest Chair by 3 May 2004. On behalf of the contest organizers, I wish you the best of luck.
Peter Lawson

2 April 2004

Updated 2 April 2004

Important Dates

  • October - November 2003: Decide on rules and data sets.
  • 7 November 2003. Deadline for Expressions of Interest.
  • 21 November 2003. Rules and Regulations Decided.
  • 24 November 2003. Submission of Abstract.
  • 5 March 2004. All preparatory tests completed.
  • 2 April 2004. Contest data set(s) released.
  • 3 May 2004. Final images submitted to contest.
  • 24 May 2004. Manuscript Submitted.
  • 21-25 June 2004. Conference Date.
The 2004 Meeting of the IAU Working Group will take place in Glasgow, Scotland, in conjunction with the SPIE International Symposium on Astronomical Telescopes (20 - 25 June 2004).
The exact date and venue of the meeting will be arranged by March 2003.