General description
The major "Vision and morphology" allows students to gain mastery in interpreting computer images. It uses the richness and diversity of the image processing applications developed at the Centre de Morphologie Mathématique (CMM). The CMM has created a theory and a method for image processing, called mathematical morphology, which has proved to be fruitful in a large variety of application areas going from quantitative microscopy to multimedia contents.
Quantitative microscopy, Industrial vision
Quantitative microscopy and medical imaging are par excellence the field of multidisciplinarity and allow to approach a great variety of sciences : medicine, biology, agronomy, geology, materials, etc. With computer vision, we get to control or surveillance activities : quality control of industrial parts, surveillance of a manufacturing process, control of a building entry, or motorway traffic monitoring.
Physical properties of materials, random model simulations
Physical modelling is the natural extension of observation and measurement. In order to study how macroscopic properties of heterogeneous media stem from microscopic properties, we use modelling and simulation techniques of probabilistic nature.
Image compression, MPEG-4 and creation of multimedia content
In the society of information that is emerging, there is no longer a separation between the production, treatment, transmission or diffusion of information. With Internet, anyone can broadcast his own TV channel. Moreover, many new services, such as remote shopping or working, use complex interactions with the content of images or video sequences. Finally, contrary to the traditional areas of image processing mentioned above, multimedia applications are characterized by the absence of a priori knowledge of the image or sequence content, which requires increasingly sophisticated tools for processing or analysis : interactive editing of audiovisual content, filtering, contouring of audiovisual objects, composition, compression, etc.
MPEG-7, indexing and content based retrieval
The huge quantity of available information, broadcasted, stored in databases, or accessible via Internet at any given time needs powerful indexing and content based retrieval techniques, capable of handling not only textual but also multimedia information. In order to achieve this, one must be able to attach meta-information to multimedia documents in order to describe their content, and to format this information according to the new MPEG-7 standard. At present, a semantic understanding of image contents is beyond computer capacities. So we have to reach a collaboration between the man and the machine to get useful and efficient indexing. Technical solutions must be designed, through fruitful exchanges between image processing technicians, film or multimedia makers, media researchers, sociologists.
Organisation of the major
Second year
In the framework of the exchanges between the Dutch ASCI Doctorate School, two GEI (Engineering schools) courses and a module given during the major period are open to Dutch PhD students. In return, our students will have a five-day course in february 2000 and another in november 2000.
The year is divided into 5 stages
- acquisition of the bases of mathematical morphology, illustrated with applications in quantitative microscopy, medical imaging and industrial vision,
- 1-week trip in Holland, to follow a course in a Doctorate School, devoted to image and MPEG coding,
- presentation of advanced techniques in segmentation, applied more specifically to multimedia sequences,
- introduction to the world of multimedia, cinema and television in particular, with the point of view of a film director and of researchers from the Institut National de l'Audiovisuel (French library of radio and television archives),
- introduction to the physics of materials and heterogeneous media. Introduction to random models and simulation techniques.
Third year
The third year is devoted to a personal research work in the field or image processing, supervised by a permanent research instructor from CMM. The months of January, May and June are fully devoted to this research usually under the form of an internship in a company. The research topics are equally distributed between the traditional areas of CMM activity: microscopy or medical imaging, industrial vision, physical properties of materials, multimedia.
Supervisors
Beatriz Marcotegui and Fernand Meyer
Centre de Morphologie Mathématique
Ecole des Mines
35, rue Saint Honoré
77305 Fontainebleau Cedex
Tel : 01 64 69 47 06
Emails : marcotegui@cmm.ensmp.fr , meyer@cmm.ensmp.fr