Variational Methods for Computer Vision

Tutorial ICCV 2011

Daniel Cremers, Bastian Goldlücke, Thomas Pock

Content

Basic concepts of variational methods: variational approach, Euler-Lagrange equations, duality, …
Foundations of convex optimization and convex relaxation techniques
Optimal first-order algorithms for convex optimization
Variational formulations for classical computer vision problems: image segmentation, optical flow, stereo and multiview reconstruction, …
Relations to discrete optimization and Markov random fields
Mumford-Shah and multilabel problems
Numerical implementation and GPU acceleration

Sunday, Nov. 6, Full day

9:30 - 10:15	Mathematical Foundations	Bastian Goldlücke
Coffee break
10:35 - 11:20	Continuous Optimization in Computer Vision	Thomas Pock
Coffee break
11:40 - 12:25	Variational Methods and Geometric Reconstruction	Daniel Cremers
Lunch break
14:00 - 14:45	Convex Relaxation for Motion and Stereo	Thomas Pock
Coffee break
15:05 - 15:50	Convex Relaxations for Multi-label Problems	Daniel Cremers
Coffee break
16:10 - 16:55	Vectorial Multilabel Problems	Bastian Goldlücke