The 2010 Interferometric Imaging Beauty Contest

TL;DR

This paper reports the results of the 2010 Optical/IR Interferometry Imaging Beauty Contest, comparing four different image reconstruction algorithms on spectral interferometric data to evaluate their strengths and limitations.

Contribution

It provides a blind comparison of four interferometric imaging algorithms on spectral data, highlighting their performance and limitations in a standardized contest setting.

Findings

RPR achieved the most accurate reconstructions.

WISARD demonstrated robustness in noisy conditions.

All algorithms showed strengths and limitations in different scenarios.

Abstract

We present the results of the fourth Optical/IR Interferometry Imaging Beauty Contest. The contest consists of blind imaging of test data sets derived from model sources and distributed in the OI-FITS format. The test data consists of spectral data sets on an object "observed" in the infrared with spectral resolution. There were 4 different algorithms competing this time: BSMEM the Bispectrum Maximum Entropy Method by Young, Baron & Buscher; RPR the Recursive Phase Reconstruction by Rengaswamy; SQUEEZE a Markov Chain Monte Carlo algorithm by Baron, Monnier & Kloppenborg; and, WISARD the Weak-phase Interferometric Sample Alternating Reconstruction Device by Vannier & Mugnier. The contest model image, the data delivered to the contestants and the rules are described as well as the results of the image reconstruction obtained by each method. These results are discussed as well as the strengths and limitations of each algorithm.