Section: Other Grants and Activities
NABLA INRIA Associate Team (with Greg Turk, Georgia Tech)
Greg Turk's team and ALICE belong to two different scientific communities: texture synthesis for Greg Turk and geometry processing for ALICE. We noticed that at a fundamental level, both scientific communities are interested in the same type of problems i.e. optimizing locally-defined criteria. In texture synthesis, the criterion is the similarity to a reference image, while in geometry processing, it is defined by discretized differential equations. In Greg Turk's team, the optimizations are performed by stochastic approaches (Monte Carlo algorithm...), while the specificity of the criteria optimized by ALICE (smoothness, convexity) allows using efficient numerical methods (Conjugate Gradient, Newton, etc.). We believe that major contributions could be achieved by combining both approaches. It can be illustrated by our first joint work: an interactive texturing tool based on stochastic approach to define the texture, and a geometry processing approach to define the mapping of this texture onto a surface. This first success leads us to believe that coupling the stochastic and numerical approaches could be made at a more fundamental level. Our ambition is therefore to define a novel unified framework to optimize complex energy functionals. Since this new framework will include both stochastic and numerical aspects, we expect important advances in geometry processing and in scientific visualization.