Having a degree in telecommunications and one in space science and engineering, the main emphasis of my work consists in designing image processing techniques to solve various problems in the fields of astrophysics and cosmology. My research mostly focuses on strong gravitational lensing, galaxy clustering and cluster lensing for the astrophysics and cosmology part. To address issues in these areas, my favourite tools are sparse optimisation, source separation and machine learning.
Strong Gravitational Lensing
Strong gravitational lensing s observed when looking at distant galaxies that lay at the behind a massive object, like a galaxy cluster or another galaxy. The massive object will act as a lens by distorting the path of any light-ray approaching it.
More about Strong Gravitational Lensing here.
Galaxy clusters are the largest known structures in the Universe. They are formed by a large number of galaxies coming in close proximity from one another ending up being gravitationally bound.
More about gravitational lensing and my activity in this area here
Sparse optimisation, which is one of my favourite tools, consists in decomposing a signal in the smallest possible number of chosen elements in order to make its reconstruction or transmission easier.
More about sparsity here
« Principal Component Analysis Robotic Deblender » is an IDL code designed to remove the central object in a postage stamp image centered on a galaxy. The aim is to recover any object in the image that is not part of the central galaxy without contamination from the foreground galaxy. The algorithm, described in Joseph et al. 2014, was used to find lensed images in the CFHTLS survey.
« Multi-band morpho-Spectral Component Analysis Deblending Tool » is a python package developed to separate objects with different colours in multi-band observations. The code, described in Joseph, Courbin & Starck 2016, assumes that images can be viewed as a linear combination of surface brightnesses emitting with different spectral energy distributions (or colours). Using this principle, we are able to separate overlapping objects with different colours by solving an inverse problem.
This technique was applied to the Hubble Frontier Fields in an unpublish work, of which you may get a glimpse here.
« Sparse Lens Inversion Tool » is a python package that inverses lensed images of galaxies while simultaneously separating the lensed image from the foreground lens light. The algorithm relies on the sparsity of the lens and source galaxies in their respective dictionaries to separate them.
Originally from a small village in the centre of France, I graduated from Telecom Bretagne (France) with a master’s degree in engineering of telecommunication and from University College London (UK) with a master’s degree in space science and engineering. Despite this very technical education, I enrolled in a PhD program in astrophysics at EPFL (Switzerland) where I applied my skills in image processing and inverse problem solving to the problems met by astrophysicists and cosmologists. In my current assignment in Princeton University, I pursue the development of deblending techniques in view of the processing of the Terra bits of soon to be brought by the LSST telescope.
My personal time goes to the various sports I practice, mostly: Badminton, Bouldering, Running, Hiking, Skiing and Fencing. I have been practising the latter for more than 20 years now. I have a first level coach degree in fencing which allows me to assist the master (main coach) in his classes and am also involved in the life of the local fencing society as a secretary. I practice mostly sportive fencing, with proficiency in all three weapons (foil, sabre and épée), but I also try to promote and practice artistic, choreographed fencing.
Today, poster presentation at the BASP frontier 2019 workshop, on how to separate images of galaxies. This method kills two birds in one stone by jointly solving… Lire la suite « BASP 2019 poster »
In astrophysics as in any other scientific field, we need to acquire data and observations of the world to build upon and to confront our predictions with.… Lire la suite « Observations in Chile: collaborative data acquisition »
I recently participated in the MT180 tournament, where contestants are asked to present their thesis in a clear, interesting and entertaining way to a public of non… Lire la suite « MT180 finalist! »
- R. Joseph
Sparse Linear Inversion for Strong Gravitational Lenses Reconstruction, Component Separation from Morphological Component Analysis with Multiple Discrimination Criteria
- R. Joseph, F. Courbin, J.-L. Starck, S. Birrer
Sparse Lens Inversion Technique (SLIT): A linear optimisation approach to the problem of lensed source reconstruction
Submitted to A&A (2017)
- R. Joseph, F. Courbin, R.B. Metcalf, C. Giocoli, P. Hartley, N. Jackson, F. Bellagamba, J.-P. Kneib, L. Koopmans, G. Lemson, M. Meneghetti, G. Meylan, M. Petkova, S. Pires
A PCA-based automated finder for galaxy-scale strong lenses
Astronomy & Astrophysics, V566, A63 (2014)
- R. Joseph, F. Courbin, , J.-L. Starck
Multi-band morpho-Spectral Component Analysis Deblending Tool (MuSCADeT): Deblending colourful objects
Astronomy $ Astrophysics, V589, A2 (2016)
- D. Paraficz, F. Courbin, A. Tramacere, R. Joseph, R.B. Metcalf, J.-P. Kneib, P. Dubath, D. Droz, F. Filleul, D. Ringeisen, C. Schäfer
The PCA Lens-Finder: application to CFHTLS
Astronomy & Astrophysics, V592, A75 (2016)
More publications here