Sparse and robust matching problem for 3D shape analysis

In this thesis we approach different aspects of the all-pervasive correspondence problem in Computer Vision. Our main results take advantage of recent developments in the emerging field of game-theoretic methods for Machine Learning and Pattern Recognition, which we adapt and shape into a general framework that is flexible enough to accommodate rather specific and commonly encountered correspondence problems within the areas of 3D reconstruction and shape analysis. We apply said framework to a variety of matching scenarios and test its effectiveness over a wide selection of applicative domains, demonstrating and motivating its capability to deliver sparse, yet very robust solutions to domain-specific instances of the matching problem. Finally, we provide some theoretical insights that both confirm the validity of the method in a rigorous manner and foster new interesting directions of research.