Universität Rostock, 2019
Abstract: The depth images from consumer depth cameras (e.g., structured-light/ToF devices) exhibit a substantial amount of artifacts (e.g., holes, flickering, ghosting) that needs to be removed for real-world applications. Existing methods cannot entirely remove them and perform slow. This thesis proposes a new real-time spatio-temporal depth image enhancement filter that completely removes flickering and ghosting, and significantly reduces holes. This thesis also presents a novel depth-data capture setup and two data reduction methods to optimize the performance of the proposed enhancement method.
doctoral thesis free access