Three-dimensional Video Generation for Realistic Broadcasting Services

In this paper, we propose a new scheme to generate multi-view video-plus-depth using a hybrid camera system, which is composed of one depth camera and multiple video cameras. In order to create the threedimensional (3-D) video, we first calculate the initial disparity for each view by projecting depth camera data onto each video camera using 3-D image warping. Then, a stereo matching algorithm is applyed to estimate the final disparity from the initial disparity in each view. Finally, we convert the final disparity into depth information in each frame to generate a multi-view depth map. Experimental results show that the proposed hybrid camera system not only provides high-qualtiy depth information for 3-D realistic broadcasting services, but also minimizes inherent problems of the conventional depth camera system.