La Vie de Josiane Zerubia: A Very Modern Woman of Science

Introducing the IEEE Fellow series

Earthzine opens an occasional series sponsored by the IEEE Foundation on IEEE Fellows with this Featured Person interview with Josiane Zerubia, who was elevated to Fellow in 2003. IEEE Fellows are “an elite global group with international recognition [who] are called upon for guidance and leadership as the world of electrical and electronic technology, continues to evolve.” Please enjoy La Vie de Josiane Zerubia: A Modern Woman of Science.

Paul E. Racette, D Sc

Editor-in-Chief

Josiane Bernadette Zerubia in Chanel frames © INRIA / Photo C. Lebedinsky

Josiane Bernadette Zerubia in Chanel frames © INRIA / Photo C. Lebedinsky

Josiane Zerubia, 51, a director of research at INRIA, the premiere French public research institute in applied mathematics and computer science, was taught at an early age by her mother Jeanne and grandmother Louise in Cannes that she could do whatever she wanted if she worked hard enough. Her mother, a hotel maid, and her suffragette grandmother, also taught her that being a girl from a poor family was not only no obstacle, it neither defined her persona nor limited her potential. “If you want to do it and work hard, you can do it – that’s how I was raised.”

In conversation from INRIA in Sophia Antipolis on the French Riviera, where the temperature in early July was as hot (32 Celsius), dry and windy as in August, she shows no sign of having forgotten that lesson.

She allowed that she needed water to drink before our interview and to the lack of air-conditioning at INRIA, but without complaint. She spoke quickly, with assurance, and in lightly accented English. That, in fact, was her only lament: she believes her fluid American English polished during her days in California has diminished because she now speaks mostly French or Pidgin English with her polyglot 25-member research group.

Her career trajectory really began at 10 years of age on a family trip to Paris when she heard a lecture on the atom at the Palais de la Decouverte, and discovered she wanted to be a researcher in physics. This aspiration continued through her high school studies of mathematics, physics, Latin, English, German and the humanities, until she got to her preparatory studies for the French “Grandes Écoles” (top ranking engineering schools) in Toulon. Although she was accepted to one of the best schools in physics and chemistry (ESPCI in Paris), she decided not to go because she quickly understood that it was necessary either to be truly exceptional, to do something original in theoretical physics, or to be very patient, to obtain results in experimental physics. She wanted results more quickly, so she switched direction to electrical engineering research at INPG in Grenoble. Having grown up without much money, her ambitions were very practical: She wanted to do research in an American company, because she thought they would be “more dynamic” than academia, and she also wanted to earn a good salary to buy a car, a TV, “basic things.”

After finishing her “Grande École” studies, she worked first for Hewlett Packard in Grenoble, even though she does not enjoy cold and snow. She found the work very challenging. “It was the early eighties, and the research lab where I worked was brand new. We were competing with IBM and others in the development of speech recognition and user-friendly interfaces for telecommunications.” She continued that work at HP Labs in Palo Alto, California, where her project leader, a professor, divided his time between HP Labs and Stanford University. He saw her passion for research and offered her a seat in Stanford’s doctoral program. Around the same time, though, HP France asked her to go back to finish the speech recognition project on which she was working (called “Cyrano”), and she decided to return to France. To her great disappointment, a few months later the project was stopped following an agreement between HP and IBM. She decided to leave HP and pursue an academic career.

Back at home by her beloved Mediterranean, she received a grant from the French government to earn her first doctorate, Doctor of Engineering in speech enhancement, from the University of Nice Sophia Antipolis. She then received an academic and research grant from IBM to earn a second doctorate, a Ph.D. in modeling and estimation for speech and image recognition, from the same institution. After graduation, INRIA awarded her a fellowship to do a post doc abroad. She had several invitations, but chose the Signal and Image Processing Institute of the University of Southern California in Los Angeles, both in order to work with an outstanding professor, and because she loves the climate there. LA’s international ambiance was also intellectually stimulating, and she made lasting friendships. She had enjoyed HP Labs in Palo Alto for the same reasons. So, when she joined INRIA in 1989, she chose to work on the Riviera in Sophia Antipolis. “I need the sun,” she said.

INRIA: the French National Institute for Research in Computer Science and Automatic Control

INRIA began in 1967 in the fertile imaginations of a small group of French academics who worked with the government of PM Georges Pompidou to insure that France would play a pioneering role in the computer revolution.

INRIA Sophia Antipolis © INRIA

INRIA Sophia Antipolis © INRIA

Today, INRIA, to quote from its website, is “the French national institute for research in computer science and control, operating under the dual authority of the Ministry of Research and the Ministry of Industry… dedicated to fundamental and applied research in information and communication science and technology (ICST). The Institute also plays a major role in technology transfer by fostering training through research, diffusion of scientific and technical information, development, as well as providing expert advice and participating in international programs.

“By playing a leading role in the scientific community in the field and being in close contact with industry, INRIA is a major participant in the development of ICST in France. Throughout its eight research centers in Rocquencourt, Rennes, Sophia Antipolis, Grenoble, Nancy, Bordeaux, Lille and Saclay, INRIA has a workforce of 3,800, 2,800 of whom are scientists from INRIA and INRIA’s partner organizations such as CNRS (the French National Center for Scientific Research), universities and leading engineering schools. They work in 150 research project-teams. Many INRIA researchers are also professors and approximately 1,000 doctoral students work on theses as members of INRIA research project-teams.”

INRIA’s research portfolio is far reaching in breadth and depth, linking basic and applied research with the private sector, and spanning computer science and applied mathematics in projects that range from the development of an artificial retina to prevention of the severe wildfires that plague the French Riviera.

In 1995, Josiane Zerubia was appointed director of research at INRIA, and in 2002 she was appointed director of research, first class,the second woman to achieve that distinction at INRIA.

la Vie a l'INRIA Sophia Antipolis © INRIA

la Vie a l'INRIA Sophia Antipolis © INRIA

ARIANA from the Greek

In 1998, Zerubia and a friend and colleague Laure Blanc-Feraud, director of research at CNRS, and mother of three children, created a research project-team. They named it “Ariana“, after the Greek goddess who fought the Minotaur. They had known each other since the 80s when they were in the same doctoral program, but had spent the next ten years working separately. When they were once again working in the same area, they said, “Voila! We did our Ph.D.s together, why don’t we create a group?” They purposely chose a female name, Zerubia said, because they wanted a powerful and inspirational female symbol. As it happened, their first hire was a female Ph.D. student, but now, eleven years later, their research group includes men and women from more than a dozen countries.

She said that one of the important aspects of this research group, which is a partnership between INRIA,CNRS, and the University of Nice Sophia Antipolis, is that the researchers put aside any historical or political differences they may have arrived with to work together as scientists, and in so doing they become friends.

It is an explicit part of her (and INRIA’s) mission to welcome scientists and students from all over the world. The French Riviera is a beautiful place to live, play (beaches, cinema, culture, skiing!) and work. The dividends go both ways: Ariana researchers currently collaborate on projects with Israel, the Maghrib, the U.S., Russia, Belarus, China, India, and Europe. As a whole, INRIA’s partnerships spread even further.

Her colleague, Renaud Marlet, Ph.D., associate research scientist at INRIA, in the SIGNES group in Bordeaux, describes her so: “On a professional point of view, I would stress her enthusiasm. I often picture Josiane as I saw her many times giving talks: she is so involved in what she presents that she kind of vibrates and her feet really seem not to touch the ground — half inch in the air out of pure inner energy. She also is a leader, able to federate people with mixed personalities.”

Automatic counting of flamingos in an aerial image. © INRIA/ARIANA (inset) and © Tour du Valat (aerial view)

Automatic counting of flamingos in an aerial image. © INRIA/ARIANA (inset) and © Tour du Valat (aerial view)

Zerubia does not consider herself a role model for other women just because she has succeeded in a scientific environment largely populated by men. She does not believe in a quota system for hiring and promotion, but in recognition through the merit of hard work and talent, although she also believes that for this to represent “equal opportunity”, people who are disadvantaged in some way should be offered help right from the beginning of their studies, both financially and educationally. She describes INRIA as a competitive workplace because there are so many talented, hard-working scientists, and as a place where she feels at home both as a director of research and as a working scientist.

She does, however, note the challenges working mothers face. She is also aware that INRIA’s efforts to reach out to school children and high school teenagers to engage them in math and science are attracting fewer girls than previously, an understandably perplexing state of affairs to a woman who became so enthralled by science as a young girl.

Another friend of many years, Catherine Carletto, a chemical engineer and marketing executive with the Unipex Group, believes Zerubia is a role model: “I am happy that this strong (in her mind) girl could be put under the light because she could be a good example for our time of a woman with good potential, working, learning and climbing to arrive at the same level as men, but without looking like a man. She has a high intellectual level, a curiosity, a rigor and a strong work ethic.”

But, Zerubia brushes aside the idea that her gender played any role in her elevation to the rank of IEEE Fellow (she was one of 43 women and 219 men who were elevated to that rank in 2003). She was “very happy” at the honor, but she entered the competition for Fellow because she was “curious” about how it worked. “I love both reflection and action” she says. “I love to be a part of the game.” She is also an Associate Editor for European Space Organizations and Industries, Space Signal and Image Processing for Earthzine, and enjoys helping to organize international conferences, publishing and other Institute activities.

Zerubia’s specialty is Markovian modeling in image processing and remote sensing. For non-mathematicians, the Encyclopedia Britannica provides this abbreviated definition: “A stochastic process is called Markovian (after the Russian mathematician Andrey Andreyevich Markov) if at any time t the conditional probability of an arbitrary future event given the entire past of the process – i.e., given X(s) for all s ≤ t – equals the conditional probability of that future event given only X(t). Thus, in order to make a probabilistic statement about the future behavior of a Markov process, it is no more helpful to know the entire history of the process than it is to know only its current state…”

Another more descriptive definition is given by Professor Oliver Ibe, University of Massachusetts Lowell, USA.

Burnt area detection after a forest fire in 2005 in Corsica, France (about 1000 hectares) SPOT 5 Image © CNES 2005, Distribution SPOT Image. Extracted boundaries © INRIA Sophia Antipolis - ARIANA

Burnt area detection after a forest fire in 2005 in Corsica, France (about 1000 hectares) SPOT 5 Image © CNES 2005, Distribution SPOT Image. Extracted boundaries © INRIA Sophia Antipolis - ARIANA

“Markov processes are used to model systems with limited memory. They are used in many areas including communications systems, transportation networks, image segmentation and analysis, biological systems and DNA sequence analysis, random atomic motion and diffusion in physics, social mobility, population studies, epidemiology, animal and insect migration, queuing systems, resource management, dams, financial engineering, actuarial science, and decision systems…”

She explained that her contribution to this field, in collaboration with her students and research partners, has been to develop Markovian models for image processing in general, and to apply them to remote sensing images in particular. The idea of limited memory in time is replaced with the idea of a “local neighborhood” in an image. The properties of an image are initially described only in each local neighborhood, but taken all together they describe the whole image: “The local neighborhood is sufficient to predict globally what will happen in the entire image.”

Ariana, the research group Zerubia still leads, uses such mathematical models to tackle inverse problems related to Earth observation and cartography. Applications of this research are legion, including using high resolution satellite images to detect and count populations of nesting flamingos worldwide, as well as penguins in Antarctica, and to detect at their onset (and assess after the fact the damage caused by) the wildfires that roar across the Riviera and Corsica every August.

In 2007, that research was made vividly immediate to her when helicopters whirled overhead dropping fire retardant on a small fire that began in a supermarket parking lot and engulfed the pines not two miles from the building where she and her group anxiously worked. The mixture of summer tourists, many of them campers, dense pine woods, high temperatures, and arid winds creates hazardous fires and urgent public safety demand for their research.

She said they obtain satellite images from several sources (the French Space Agency, CNES; the French National Geographic Institute, IGN; the French National Forest Inventory, IFN; the German Space Agency (DLR); Thales Alenia Space; EADS Astrium;…), and work with fire brigade members to draw maps that enable them to forecast the probable risk and direction of new fires, and to analyze burned areas for insurance purposes and the like.

Examples of past ARIANA projects: Agricultural and  peri-urban areas classification result, using Markov random fields from a hyperspectral image. Left: © Astrium/EADS. Right: ©  INRIA/ARIANA

Examples of past ARIANA projects: Agricultural and peri-urban areas classification result, using Markov random fields from a hyperspectral image. Left: © Astrium/EADS. Right: © INRIA/ARIANA

Gazing On the Future of Earth Observation

The developments in remote sensing over the last twenty years have made the challenges of observing and mapping the Earth more exciting for her.

“Twenty years ago,” she said, “with satellite images from Landsat in the USA or SPOT 1or 2 in Europe, we could analyze images of cities, but it was difficult to see any detail. Now, we can count the buildings and see the details of trees, rivers and roads – even the cars on the roads. So, we have to change the mathematical models.”

“Before, when we had lower resolution, we could use Markov random fields. Now, with very high resolution images, we have to introduce geometry in a stochastic way into these models. This is why we have introduced marked point processes.”

The research of the Ariana group will continue to build on this work – improving the quality of the images obtained from remote sensing, and introducing geometry into the models (using not only stochastic processes, but also variational methods) in order to extract more detailed information from the images.

She is optimistic that environmental efforts will be in time and sufficient to rescue planet Earth. “I think awareness of what we have done to the environment is growing and that what we do has serious consequences to Nature. I hope the new generation will be more careful. I am not a pessimistic person. I am hopeful.”