Skinning a Deformation Model

(Semester Project)

Virtual Reality Lab

Semestre




Description:

The CO-ME, Computer Aided and Image Guided Medical Interventions, is an NCCR long-term project that involves many Swiss Institutes and Universities. The aim of the project is to use the potential of information technology for the optimization of medical interventions in order to improve both the treatment of individual patients and overall health care for the whole society.

Our lab is contributing to the project by creating a biomechanical model of joints. In this model, joint postures and contact between different tissues are closely related to allow future applications to study normal and pahological cases of motion.

The problem

We are working on a physically based deformation model of soft tissues to represent biological tissues like cartilage and ligament. Such tissues are being modeled based on a mass-spring like model called molecular model. In this model, spherical molecules are combined together to compose the volume being represented. See images.

The surfaces of objects modeled with this approach are not smooth. Because of the spheres, these surfaces are irregular, which causes the simulation of contact to be non-realistic for many applications. In addition, the visual aspect of such objects cannot be satisfactory realist because of its wrinkled surface.

Thus, the problem is to put a smooth surface model (a kind of skin) on the top of the deformation model described above that follows the inner model deformation.

Student’s work
The information available are the original triangular mesh and the molecular model buit from it. The student should first study existent skinning methods and decide either to use the original mesh, anchoring it on the molecules and deforming it accordingly, or implement a continuous surface model (splines, bezier, etc.) and use the molecules positions as control points.

Then, he/she is asked to use and update the existent C++ classes and methods, and implement a new set of methods to fit the surface on the existent deformation model. Methods based on V-Collide are already available within the existent classes to compute collision detection. They should be revised and improved if possible and/or necessary to integrate to the new surface model. The evaluation of results will be focused on the visual and performance aspects.



Platform:

PC, C++.

Prerequisite:
Good knowledge in C++.

References :
CO-ME : http://co-me.ch
Deformation model paper: http://vrlab.epfl.ch/Publications/pdf/Maciel_Boulic_Thalmann_IS4TM_03.pdf
Diploma project on force-feedback http://vrlab.epfl.ch/~amaciel/projects/obuchwalder2003.pdf

Note: Project can be done either in English or French.

Responsibles: Anderson Maciel, Patrick Lemoine, Sofiane Sarni  Tel: 5248,  Room: INJ 118, Dept: DI
Emails: Anderson.Maciel@epfl.ch; Patrick.Lemoine@epfl.ch; Sofiane.Sarni@epfl.ch