Image reconstruction algorithms

WP 4.3: Image reconstruction algorithms

This sub-WP will adapt the best of the existing image reconstruction methods to the new types of data provided by the multi-telescope interferometers. To make the algorithms more user-friendly, specific R&D will be undertaken to simplify the control of the algorithms.

Whatever the on-line help provided by a software interface, a minimum level of knowledge is required to choose the most appropriate algorithm parameters and perform the best reconstruction for a given data set. We will therefore start by writing a background document describing the work flow of image reconstruction to explain its fundamental principles to the users. This document will serve to unify the methodologies being used and help to specify interfaces and to schedule the algorithm developments.

Means to account for the new types of data will be studied and implemented based on the description of the data generated by WP4.2. These studies will include: hyper-spectral image synthesis from multi-wavelength interferometric data which will take into account off-axis and on-axis phase referenced data and image reconstruction with a variable PSF. The deliverables will be either modified algorithms based on the existing state of the art (BSMEM, MiRA and Wisard) or recipes and case studies using existing algorithms (such as Airy-LN) which can best be used to obtain useful images from instruments such as Matisse, Gravity, LINC-Nirvana, VEGA/CHARA and Pionier.

To fully exploit the know-how of each partner (some of them are the authors of the above-mentioned algorithms), it is not sensible to work on a single implementation so a few different algorithms will be developed or improved. We will however pool as much work as possible. In particular, the tools produced will deal with the same input and output file formats (OI-FITS for the interferometric data and multi-dimensional FITS for the multi-wavelength images and other data) and use similar parameters to drive the reconstruction (e.g., pixel size, regularization method and hyper-parameters). Finally whatever the programming language used for their implementation, the different algorithms will have a similar command line interface to allow them to be driven with a common graphical interface. The specifications of this interface are part of WP4.4.