Recent results on energy minimization via Graph Cuts have opened the door to a series of advances in low-level image problems such as segmentation, restoration, classification, stereo, and multi-camera scene reconstruction. By exploiting this new energy minimization results, this project aims at conceptual and algorithmic developments in the field of Phase Unwrapping (PU), namely, the following:
1. Design exact minimizers for Lp-norm based PU, for p >= 1
2. Design sub-optimal minimizers for Lp-norm based PU, for
0 < p < 1
3. Design new robust clique potentials tailored to PU, based on the Bayesian paradigm
4. Design PU schemes taking advantage of diversity (e.g., space or frequency), i.e., of more than one sources of observed wrapped phase.
1. Develop, in C/C++, efficient Min-Cut/Max-Flow algorithms taking advantage of the graph structure inherent to the problems listed above
2. Develop efficient algorithms implementing the minimizers listed above
3. Develop “real world” application-oriented prototype demonstrators for these algorithms.
|Start Date: 12-12-2005|
|End Date: 12-12-2007|
|Team: José Manuel Bioucas Dias, Paulo Alexandre Carapinha Marques|
|Groups: Pattern and Image Analysis – Lx|
|Partners: Dr. Mario Ries from “Institut des Neorosciences de Bourdeaux|
|Local Coordinator: José Manuel Bioucas Dias|
|Links: Internal Page|