Prof. Rachid Deriche, Research Director at INRIA Sophia Antipolis, was awarded the 2013 French Academy of Sciences Grand Prize of the EADS Corporate Foundation in Computer Science. This award recognises the achievements of a scientist in a French laboratory who has made exceptional contributions to the vitality and influence of computer science research while building outstanding cooperation with industry. Inspire Magazine spoke with Prof. Deriche to find out more about his research activities that led to this prize, which has been officially awarded at the Institut de France on the 15th of October 2013.
Inspire Magazine: Prof. Deriche, many thanks for speaking to Inspire Magazine and its readers, and many congratulations on winning this prestigious award. How do you feel about this achievement?
Rachid Deriche: Thanks for your congratulations and also for your interest to know more about my work. I’m very proud and happy and I thank the Academy of Science and the EADS foundation for this prestigious award and would like to also thank my institute INRIA for all the facilities and the freedom I had to perform my research. I would like to dedicate this prize to my family, my colleagues and my PhD students with whom I have closely worked over all these years. Last but not least, I would also like to dedicate this prize to all my former professors and to all my friends in Algeria, in particular those with whom I have been at the Ecole Polytechnique in Algiers almost 40 years ago and the colleagues with whom I am currently collaborating in Algeria.
IM: Could you tell us a little bit about your background and your academic journey that led you to where you are now?
Rachid Deriche: I am 59 years old and was born in Thenia, 50 kilometers east of Algiers where I lived till I joined the high school Amara Rachid in Ben Aknoun, Algiers. Right after my high school diploma, the Baccalaureate, in Mathematics in 1973, I joined the Ecole Nationale Polytechnique of Algiers in El Harrach where I studied electronics engineering till my 4th year. In 1977, I joined the Ecole Nationale Superieure des Telecommunications in Paris from which I graduated in 1979. I then received the Ph.D degree in Mathematics from the University of Paris IX, Dauphine in 1982 and the Habilitation à Diriger des Recherches (HDR diploma) from Nice Sophia Antipolis University in 1991. I did my Ph.D thesis at INRIA, the French National Institute in Computer Science and Control in image processing, computer vision and neuro-imaging, I did this in its research center at Rocquencour, Yvelines, close to Paris, till 1988 and then in its research center located at Sophia-Antipolis, close to Nice in the French riviera. So I started to contribute to the fields of image processing in the early 80’s and then moved to the domain of computer vision in the early 90’s before shifting my research interest to the domain of neuro-imaging in the early 2000’s.
Rchid Deriche: I am currently Research Director at INRIA Sophia Antipols – Mediterranee where I lead the research activities of the Athena Project Team with the objectives to explore the Central Nervous System (CNS) using computational imaging. We focus on signal and image recording from Diffusion Magnetic Resonance Imaging (dMRI), Magneto-Encephalography (MEG) and Electro-Encephalography (EEG). Our goal is to develop rigorous mathematical models and computational tools for acquiring, modelling and analysing the complex CNS anatomy and function. These models and tools will help us to better understand the architecture and the functioning of the human CNS. Exploring innovative directions to solve these challenging problems will push forward the state-of-the-art in Computational Anatomical and Functional CNS Imaging and in a long term, will help our collaborators with hospitals to use our expertise to address important clinical and neuroscience questions.
IM: You also work closely with the industry?
Rachid Deriche: Yes indeed, the strategic objectives of INRIA is to advance the digital society, directed towards scientific excellence and technology transfer. Therefore, combining scientific excellence with technology transfer has always been and is still at the heart of all my contributions. I served as a Principle Investigator in many European and national projects since the mid 1980’s and within all these projects, I had numerous collaborations with industrial companies and Small and Medium Enterprises (SMEs). Currently, I am closely collaborating with a very promising SME, Olea Medical, which provides innovative software solutions to healthcare professionals worldwide. I am also a scientific consultant for the International Technological Group Safran as well as Olea Medical and I used to be a consultant for other companies and SME’s like Realviz, Poseidon and IFP.
IM: What are your main research activities and how have these impacted related industry?
Rachid Deriche: My initial research activities were first on image processing and computer vision, with a particular emphasis on early vision, on the geometry of multiple views for 3D reconstruction and on variational approaches, partial differential equations (PDE’s) and level-sets. I co-founded the start-up Realviz in 1998, it has been extremely successful in postproduction and special effects, which illustrates the impact of the results I produced. Realviz’s technology provides efficient ways to generate 3D content and visual effects from photo imaging and 2D environments. Its products are mainly based on the software my colleagues and I developed at INRIA for panoramic photography, image-based modeling, match moving and optical motion capture. In 2008, Autodesk has acquired all the assets of Realviz.
In 2002, following the successful technology transfer to Realviz, I shifted my research interest to computational neuroscience and neuro-imaging. This thematic mobility towards a totally new research domain was very risky but scientifically challenging and extremely exciting. In this new research domain, my research work primarily focused on computational brain imaging with a particular emphasis on the understanding and the computation of the anatomical connectivities in the human brain through Diffusion MRI and its possible combination with other imaging modalities such MEG or EEG. For the recent years, I have been mainly active in developing pioneering algorithms for the analysis and clinical application of Diffusion MRI data.
The in-vivo and non-invasive imaging modalities we use raise a large amount of mathematical and computational challenges. I am strongly involved in all theses quests, particularly through a close collaboration with national and international laboratories and hospitals. These long term and ambitious challenges, when successful, will be of great help to make true the dream to continue developing innovative research and effectively contribute to reduce the number of people suffering from CNS diseases, and this work is done in collaboration with Olea Medical.
IM: Put together, these endeavours essentially led to your award of the EADS Grand Prize in Computer Science. Could you go into more details about the research contributions behind this collection of works?
Rachid Deriche: My contributions are structured on three research axis, all based on mathematically well-founded and robust methods. Mathematical modelling, accurate analysis and efficient computation are at the heart of almost all the common mathematical foundations and frameworks I proposed to address the crucial criteria of robustness, well-posedness, accuracy, and computational efficiency, just to name a few.
The first research axis I developed was related to the domains of Early Vision and 3D vision with a particular emphasis on image features extraction and the geometry of multiple views for 3D reconstruction. In the Early Vision domain, I have mainly contributed to develop fast algorithms for low-level vision and efficient filtering techniques to optimally detect edges, corners and other important image features. All these contributions are extremely well referenced and are, even years after, the state-of-the-art techniques in early vision. Interestingly, some of them have been adapted to some specific real time applications using DSP, FPGA, ASIC etc. In the 3D Vision domain, my contributions have permitted to develop a whole set of tools issued from the projective geometry to process and analyse images issued from multiple cameras or from temporal sequence of images. Among the most relevant works, I highly contributed to introduce the well-known fundamental matrix and to develop its first robust estimation algorithm for matching two uncalibrated images. These contributions have been a breakthrough and opened new perspectives in the computer vision community.
I also contributed to the development of the first algorithm on mosaicking, at the heart of the start-up Realviz’s flagship product Stitcher software for the creation of panoramas and 360°virtual tours. Also, worth to mention, are my pioneering work on tracking using Kalman Filtering and my contributions to the camera self-calibration problem using simplified Kruppa’s equation through the singular value decomposition of the fundamental matrix.
My main contributions within my second research axis have been on pioneering and developing original techniques based on variational methods, partial differential equations and level-set methods. I have started this research axis in the mid 1990’s and since then, I have been interested in all aspects of these techniques, including mathematical modelling, well-posedness analysis, efficient algorithms and contributed to apply them in a multitude of applications. I have obtained a whole set of theoretical as well practical results in scalar image regularization, image sequence processing and in mathematical analysis of the optical flow. Since then, a generalisation of these fundamental contributions to images lying in a more general manifold, to vectorial and multi-valued images with/without constraints such as color and tensor images have been developed with great success. Then I particularly contributed to solve the problem of image segmentation using geodesic active contours and regions through level-set techniques. After the great success obtained in detecting and tracking moving objects using geodesic active contours, I proposed a totally innovative framework, the geodesic active region, which has been successfully adapted and applied to important problems such as scalar image segmentation, color and textured images and multivalued images detection, tracking and motion estimation. More recently, I have extended this variational framework to process 3D volumic medical images of tensors and Orientation Distributions Fields.
Finally, my third research axis started in 2002 following my decision to shift my research interests to computational neurosciences and neuro-imaging. My contributions have been mainly focused on the computational Diffusion MRI (dMRI) field. This has led to fundamental, theoretical, methodological and practical contributions to Diffusion Tensor Imaging (DTI) and High Angular Resolution Diffusion Imaging (HARDI) with original and state-of-the-art algorithms in estimation, regularization, segmentation, tractography and clustering in DTI. I have mainly introduced and advocated the use of new tools and concepts from Riemannian Geometry to efficient DTI and HARDI processing and developed a whole set of state-of-the-art algorithms in dMRI that include second as well as high order tensor and HARDI models. More recently, I started to tackle important applications such as optimising dMRI acquisitions using Kalman filtering tools and Compressed sensing theory, automatically clustering fibre paths to facilitate group-based statistical analysis and extend the Riemannian framework developed for the second order to high order tensor models, Ensemble Average Propagator (EAP) and Orientation Distribution Function (ODF) computing. All these contributions, most of which are extremely well referenced, are available in my web page.
IM: Why is cooperation between academic research and industry so important?
Rachid Deriche: At INRIA, we do both theoretical and applied research in the computing and mathematical fields of digital sciences and INRIA’s missions are to produce both outstanding research and to ensure the impact of its research on the economy and society in particular. Knowledge and technology transfer are extremely important and can be performed through different ways: knowledge transfer can be performed through actions such as teaching and directing Ph.D thesis for instance. That’s why I am Adj. Director of the STIC Doctoral School at Nice Sophia-Antipolis University and also teaching graduate courses on Biomedical Imaging, Computer Vision and Image processing in the Master of Science in Computational Biology and in the engineering School Telecom Sud Paris. I graduated and co-graduated 40 PhD students, most of them have prestigious positions in the academic or industrial world. Technology transfer can be performed through various actions: direct collaborations with industry and SME’s, creation of Start-ups and/or consulting activities. These actions are complementary and helpful to tackle very interesting and challenging problems, that if solved could have an important social and economical impact. This is what happened for me while trying to bridge the gap between the academy and the industry and this is why I consider that a solid cooperation between academic research and industry is important: it clearly helps to push the frontiers of knowledge, and acts without any doubt as an important driving force for innovation and economic growth.
IM: What is your perspectives on the interaction between industry and academia in Algeria and how can this interaction be leveraged?
Rachid Deriche: Unfortunately, it’s very difficult for me to answer this complex question mainly because I do not know the situation of the R&D in the industry in Algeria and do not know if there is a framework that has been implemented to push forward more interactions between industry and academia. What I can say based on my experience, is that I have already been invited to give some lectures and seminars in Algeria, and during each visit, I have always been very impressed by the motivation, the thirst of knowledge and learning of the students I met. It seems to me that this human resource is the best force on which to invest and the best support lever on which a solid interaction between industry and academia should be grounded. For instance, we could think, if this does not already exist, about an industrial agreement of training through research, to support R&D projects with internships and Ph.Ds granted by Industry. However, and as I mentioned above, I don’t have a clear idea of how things are locally organized, I think the best answer to your question will certainly have to come from colleagues in Algeria.
IM: Many thanks again for taking the time to speak to Inspire Magazine and all the best with your future endeavours.
Rachid Deriche: Thank you very much. It has been a pleasure to answer your questions. All my best wishes to you and your scientific magazine.