A statistical model of Riemmannian metric variation for deformable shape analysis

Non-rigid transformations problems have been largely addressed recently by the researchers community due to their importance in various areas, such as medical research and automatic information retrieval systems. In this dissertation we use a novel technique to learn a statistical model based on Riemmannian metric variation on deformable shapes. The variations learned over different datasets is then used to build a statistical model of a certain shape that is independent from the pose of the shape itself. The statistical model can then be used to classify shape that are not present in the original dataset.