Hand Pose Estimation from RGBD Images

Abstract

It is surprising that even with increasing ubiquitousness of Augmented Reality applications on mobile devices, users nowadays still interact with said applications via on-screen controls, rather than controlling presented environment directly with their hands in real-world. To enable this degree of interactivity, a stable and robust hand pose estimation pipeline is needed. This thesis therefore serves as a study of a possible approach to hand pose estimation that consists of two parts; segmentation and estimation of hand model parameters.