Direct access to content


French version


Home > Research > Signal & Image Processing

SIP : Signal & Image Processing

Imaging and image analysis perfuse nowadays all sciences, and technologies from defense to entertainment, from medicine to remote sensing. CMLA's research leads to effective and verified applications of mathematics to biology, cytology, medical diagnose, anatomy, satellite imaging, satellite design, computer graphics, camera design, medical imaging, industrial inspection, and defense. Imaging is booming and CMLA is one of the conspicuous actors of this explosion.

Founded by Yves Meyer (Gauss prize 2010) and Robert Azencott, the CMLA Image Group reached worldwide fame :
  • by the invention of wavelets and the establishment of the JPEG 2000 image exchange standard based on wavelets,
  • by the development by Alain Trouvé and Laurent Younès of a computational anatomy based on differential geometry (the so-called Diffeomorphometry),
  • by the breakthroughs of Mila Nikolova (Prix Monpetit 2010) on nonsmooth optimization for imaging,
  • by the invention of new Earth observation satellites by Bernard Rougé in collaboration with CMLA : Pleiades (now flying), SMOS and SMOS-Next 
  • and by the invention by Toni Buades, Tomeu Coll and Jean-Michel Morel (ERC laureate) of nonlocal methods for image processing.

The long list of industrial and institutional partners and agencies supporting or collaborating through contracts with CMLA on images includes Centre National d'Etudes Spatiales, DxO Labs, Institut Pasteur, European Research Council, US Office of  Naval Research, Direction Générale de l'Armement, Electricité de France, and several spinoffs...


  • Jean-Michel Morel : Clay Senior Scholar 2010, elected 2010 Senior Member of Institut Universitaire de France, editor in chief of Image Processing on Line and of SIAM Journal on Imaging Science (second highest rated journal among 241 applied mathematics journals), INRIA-Académie des Sciences Grand Prix 2013...
  • The 2011 paper "Scale Space Meshing of Raw Data Point Sets" is distinguished at the Tenth Anniversary of the Symposium on Geometry Processing. Plenary conference by Julie Digne on the theme of this paper. 2012: Best PhD award for Julie Digne, Fondation Mathématique Jacques Hadamard.
  • 2012: The affine invariant image recognition method ASIFT (invented by Jean-Michel Morel and Guoshen Yu) wins the Chinese-French competition for entrepreneurship and, "Grand Prix" among 48 competitors.
  • 2013: Jean-Michel Morel invited to a lecture at International Mathematical Conference (ICM 2014) in section 17: Mathematics in Science and Technology.
  • 2013 : The "denoising clinic" integrated in DxO Pro image processing software.


Jean-Michel MOREL
IUF Senior Member
Grand Prix INRIA Académie des Sciences 2013
CNRS Innovation Medal 2015
IEEE Longuet Higgings Prize 2015

Mathematics of Stereoscopic Satellite Images

The MISS project embodies the partnership between the CMLA and the CNES, started in 2007. CMLA coordinates mathematical and numerical research needed towards the implementation of a stereo pipeline for low-baseline satellite images and for stereo reconstruction.

MVA Master Program

The research master MVA (for Mathematics, Vision, Learning) founded by Robert Azencott is by now considered by experts the best master on imaging and learning theory worldwide. More...

Image Processing On Line

logo IPOL
IPOL is an open access peer reviewed online journal
& demonstrator in image processing: integrating software development in a new publishing model.
See latest papers...
Subscribe online...

IPOL is a recent interaction and research tool for image processing and computer vision. IPOL is about to remodel the way the image processing and image analysis community works, by documenting completely all algorithms, including code and online executions, peer-review of code, and establishment of a reliable state of the art.

Projet Twelve Labours in Image Processing

"Twelve Labours in Image Processing" is funded by the ERC (European Research Council) and aims at providing to the research community and the industry with the first validated complete image processing chains, by establishing or reconsidering the basic algorithmic atoms of this discipline. The main underlying assumption of this endeavor is that most image processing and analysis algorithms are actually universal, applicable to all digital images after the adequate mathematical modeling.
  • Key words : Raw image denoising ; Demosaicking ; Mathematical camera calibration ; Subpixel accuracy ; False alarm control ; 3D Reconstruction