Opticon FP7 Board Meeting 29-30 Oct. 2014, Granada

Interferometric Image ReconstructionWork-package of Opticon FP7

Éric Thiébaut (CRAL), Paulo Garcia (Univ. Porto),John Young (Cambridge), Jörg-Uwe Pott (MPIA Heidelberg),

Guillaume Mella (IPAG/JMMC), Gilles Duvert (IPAG/JMMC), Jean-Baptiste le Bouquin (IPAG/JMMC), Martin Vannier (OCA), Thibaut

Paumard (LESIA)

Opticon FP7 Board Meeting 29-30 Oct. 2014, Granada

Image reconstruction project for OPTICON FP7

Rationale and context

– with 4 to 6 telescopes recombiners, 2nd generation VLTI instruments and recent optical interferometers (e.g., CHARA) are targeted at multi-spectral imaging (R ~ 104, Δθ ~ 10-3 ”)

– for the scientific returns of these instruments, tools for image reconstruction usable by non-expert astronomers are required

– image reconstruction algorithms for interferometry (BSMEM, WISARD, MiRA, etc.) are mature but:

● require substantial expertise

● mostly provide monochromatic image reconstruction

● lack of documentation

● not necessarily freely available

Summary of the project

– make the R&D and tutorial/software development to provide image reconstruction algorithms from optical interferometry data to general users

Giant star surface from simulated NPOI data (Berger et al., 2012, A&A Rev. 20, 53)

Opticon FP7 Board Meeting 29-30 Oct. 2014, Granada

Work-packages

● WP1 – Project management (CRAL)

● WP2 – Data samples and astrophysical model images

– model images of astrophysical objects (FEUP)

– synthetic data for LINC-Nirvana, Gravity and Matisse (MPIA, LESIA, OCA)

– real data from Amber, LINC-Nirvana, Vega/Chara and Pionier (IPAG, MPIA, OCA)

● WP3 – Image reconstruction algorithms

– unified image reconstruction description (UC)

– algorithms derived from BSMEM, MiRA and Wisard (UC, CRAL, OCA)

● WP4 – User interface and user guides

– algorithm interface specification (CRAL)

– graphical user interface (JMMC)

– tests and benchmarks (JMMC)

– documentation and cookbooks (FEUP)

Opticon FP7 Board Meeting 29-30 Oct. 2014, Granada

Partnership● Lyon: CRAL (Éric Thiébaut)

– image reconstruction (MiRA)

● Univ. Porto (Paulo Garcia)

– sciences cases, cookbooks, tests

● Univ. Cambridge (John Young)

– image reconstruction algorithm (BSMEM)

● MPIA Heidelberg (Jörg-Uwe Pott)

– LINC-Nirvana (LBT) case

● IPAG/JMMC (Gilles Duvert, Guillaume Mella, Jean-Baptiste le Bouquin)

– data from PIONIER and AMBER

– graphical user interface

● OCA (Martin Vannier)

– data from VEGA, simulations (MATISSE)

– algorithm

● LESIA (Thibaut Paumard)

– simulation and test cases (GRAVITY)

Opticon FP7 Board Meeting 29-30 Oct. 2014, Granada

First deliverable: Unified Image Reconstruction Description

● objectives and contents

– formal description of image reconstruction in optical interferometry;

– general: all considered algorithms (BSMEM, WISARD, MiRA, etc.) can be expressed in this framework;

– required to homogenize and unify the inputs and outputs of these algorithms;

– didactic: give background information needed for the end users to understand the principle of image reconstruction (this knowledge is needed to properly use the software);

● document written by John Young (Univ. Cambridge) and Éric Thiébaut (CRAL)

● delivered on March 2014

● minor amendments expected

– new OIFITS2 format

– feedback from others

● is it worth being published?

Opticon FP7 Board Meeting 16/10/2012

Image Reconstruction in a Nutshell

positivitynormalization

strict constraints (feasible set of acceptable solutions):

basis functionsparameters

direct model:N = number of pixelsL = number of spectral channels

criterion for reconstruction:

likelihood (enforcesagreement of model

with data)

regularization(enforces prior

information)

hyper-parameter

beware: lots of parameters (as many as pixels), non­linear, non­convex

Opticon FP7 Board Meeting 29-30 Oct. 2014, Granada

Optical Interferometric Data: OIFITS-2

● OIFITS = file format to exchange optical interferometric data, widely used but:

– some ambiguities (not everyone understand the information in the same way)

– some limitations (only calibrated data, not suitable for differential phase, etc.)

– not suitable for new ESO instruments (GRAVITY and MATISSE) and probably others

● OIFITS-2

– fix these issues and add needed features

● provide spectral flux data● polarimetry● raw and calibrated data

– important for the JRA to contribute, mainly:

● John Young (Univ. Cambridge): description, testing of the new format, provide OIFITSlib, an open source library (https://github.com/jsy1001/oifitslib) for reading/writing/filtering OIFITS data files with preliminary support to OIFITS-2

● Gilles Duvert (JMMC/IPAG): description and formalization of the new format

– new format specification will be soon published (99 citations for «old» OIFITS paper)

Opticon FP7 Board Meeting 29-30 Oct. 2014, Granada

Fizeau Interferometry: LBT-LINC

● sub-workpackage led by MPI Heidelberg (Jörg-Uwe Pott):

– test cases and software for Fizeau interferometric mode of the LBT (LINC-NIRVANA)

– Rainer Köhler hired on the WP4 money on Jan. 2013

● unfortunately:

– LBT-LINC interferometer is not yet open for the community, although they want that this eventually happen

– LBTI: mid-infrared on-axis recombiner with narrow field of view, operating in speckle mode, but not really "long-baseline" interferometry

– after some preliminary work on LBT-LINC (simulation + reconstruction from u-v plane data) it seems more reasonable and productive to contribute to Matisse (Köhler, SPIE 2014)

→ needs to synchronize with what is done around the Matisse case at CNRS/Nice

Simulated image of M87 Recombined imageObserved LINC images at hour angles -45°, 0° and 45°

Opticon FP7 Board Meeting 29-30 Oct. 2014, Granada

Image Reconstruction Algorithms

● Existing algorithms

– BSMEM (John Young, Univ. Cambridge)

● now free and easy to install (used by the students of the 2013' VLTI Summer School in Barcelonnette);

– Wisard (Martin Vannier, Lagrange & Gilles Duvert, JMMC/IPAG)

● now freeware, ported to GDL (GNU Data Language, an opensource version of IDL™)

● a number of issues have been fixed;

– MiRA (Éric Thiébaut & Ferréol Soulez, CRAL)

● Yorick version freely available (used by the students of the 2013' VLTI Summer School in Barcelonnette);

● with Ferréol Soulez (POLCA project) to simplify the image reconstruction for the end user and investigate Monte-Carlo strategies to achieve global optimization.

● New algorithms for true multi-spectral image reconstruction (link with POLCA project)

– Painter (Schultz et al., 2014, J. Opt. Soc. Am. A 31, 2334)

– MiRA 3D (Soulez et al., 2014, SPIE)

– successful but difficult: probably not for the end-user

Opticon FP7 Board Meeting 29-30 Oct. 2014, Granada

Image Reconstruction Algorithms

● Existing algorithms

– BSMEM (John Young, Univ. Cambridge)

● now free and easy to install (used by the students of the 2013' VLTI Summer School in Barcelonnette);

– Wisard (Martin Vannier, Lagrange & Gilles Duvert, JMMC/IPAG)

● now freeware, ported to GDL (GNU Data Language, an opensource version of IDL™)

● a number of issues have been fixed;

– MiRA (Éric Thiébaut & Ferréol Soulez, CRAL)

● Yorick version freely available (used by the students of the 2013' VLTI Summer School in Barcelonnette);

● with Ferréol Soulez (POLCA project) to simplify the image reconstruction for the end user and investigate Monte-Carlo strategies to achieve global optimization.

● New algorithms for true multispectral image reconstruction (link with POLCA project)

– Painter (Schultz et al., 2014, J. Opt. Soc. Am. A 31, 2334)

– MiRA 3D (Soulez et al., 2014, SPIE)

– successful but difficult: probably not for the end-user

Herbig Be HD98922, PIONIER data in H band, MiRA-3D algorithm (Soulez et al., 2014, SPIE)

spectrally integrated image spectral slope ↔ temperature

Opticon FP7 Board Meeting 29-30 Oct. 2014, Granada

Software Development

● a number of issues:

– exchanging data → OIFITS-2

– making algorithms freely available (done)

– making software easy to install (C/C++, IDL/GDL, FORTRAN, Yorick, MatLab/Octave)

→ we shall improve on this (or provide and maintain precompiled versions)

– making software easy to use → that's R&D in progress

● work undergone:

– try to share software (seems obvious idea but not easy as softwares already exist)

● John Young: OIFITSlib an open source library (https://github.com/jsy1001/oifitslib) for reading/writing/filtering OIFITS data files with preliminary support to OIFITS-2

● at CRAL: massive rewrite of all our large scale optimization code (notably VMLMB which is also used in Wisard) → portable C/C++ and Java

– build portable code:

● at CRAL, with Jonathan Léger (just hired on OPTICON funds): development of a Java toolkit for inverse problems and image reconstruction (https://github.com/emmt/TiPi)

● performances surprisingly good but require to make a Java version of NFFT

Opticon FP7 Board Meeting 29-30 Oct. 2014, Granada

Simulations and Synthetic Data

● different types of objects are considered

– LESIA (Thibaut Paumard)

● development of a simulation code (GYOTO: https://github.com/gyoto/Gyoto) for computing orbits and ray-traced images in general relativity (Vincent, Paumard, Gourgoulhon & Perrin, 2011, «GYOTO: a new general relativistic ray-tracing code», Classical and Quantum Gravity 28, 225011 [arXiv:1109.4769]).

● objective is to simulate Sgr A* and its close environment

● will be used to generate synthetic data representative of GRAVITY observations

– Univ. Porto (Paulo Garcia)

● point sources and extended objects

● comparison of phase closure vs. phase referenced data

● reconstruction with MiRA

● rules of thumb to derive the regularization level and estimate the effective angular resolution

→ we need an image quality criterion

– MPIA (Jörg-Uwe Pott)

● Matisse case: close environment of stars (disk, dust, etc.)

● reconstruction with MiRA

Opticon FP7 Board Meeting 29-30 Oct. 2014, Granada

Tasks, Deliverables and Roadmap

Deliverables Name Completed Schedule(months)

Model images of astrophysical objects 25% 6

Synthetic data for Gravity 50% 18

Synthetic data for Matisse 50% 18

Training data from Pionier 50% 18

Data from Vega/Chara 50% 18

Synthetic and real data from LBLT-LINC x 30

4.1 Unified image reconstruction description 100% 12

Algorithm interface specifications 25% 18

Prototype of algorithm derived from BSMEM 50% 30

Prototype of algorithm derived from Wisard 50% 30

Prototype of algorithm derived from MiRA 50% 30

Data reduction tools for Fizeau interferometry x 30

4.3 Graphical user interface 0% 48

Test and benchmark results 0% 48

4.2 Documentation and cookbook 0% 36