Speaker: Nicolas Charon
Abstract. This talk intends to motivate the long-standing interest in some concepts borrowed from geometric measure theory within the field of shape analysis and its applications to computational anatomy. I will introduce a synthetic notion of generalized distributions known as oriented varifold that provides a parametrization-free representation for shapes and explain how it leads to an effective framework for building simple distances on spaces of curves or surfaces. I will then present a few examples of applications to different problems like diffeomorphic registration or shape clustering and statistics.