Title: PDE's and Level Set based techniques for Vector valued images regularization and dynamic segmentation : Frameworks and Applications



			   Rachid Deriche
     http://www-sop.inria.fr/odyssee/team/Rachid.Deriche/index.en.html
                  Inria Sophia-Antipolis - France


Abstract:
This talk will describe some of the problems that I have recently addressed 
to solve important image processing and computer vision applications. Two 
variational frameworks that lead to apply Partial Differential Equations and 
Level Set based techniques to regularize and dynamically segment vector 
valued images will be presented and discussed. In the first part of my talk, 
I'll address the problem of vector-valued image regularization and its 
applications to a wide variety of image processing problems, including in 
particular Diffusion tensor MRI images regularization and flow visualization. 
A second part will then be devoted to show how level set based techniques can 
be successfully used to segment vector valued images while integrating 
different cues such as color and texture for instance. Promising results 
obtained on various real images will illustrate the proposed solutions.