Search

International audience; With the rapid emergence of various 360 video capturing devices and head-mounted displays, providing immersive experience using 360 videos is becoming a topic of paramount interest. The current techniques face motion sickness issues, resulting in low quality of experience. One of the reason is that they do not take advantage of the parallax between the divergent views, which may be helpful to provide the correct view according to the user's head motion. In this paper, we propose to get rid of the classical ERP representation, and to synthesize arbitrary views using different divergent views together with their corresponding depths. Thus, we can exploit the parallax between the divergent views. In this context, we assess the feasibility of the depth estimation and the view synthesis using state-of-the-art techniques. Simulation results confirmed the feasibility of such a proposal, in addition to possibility to achieve sufficient visual quality for a head motion up to 0.1m from the rig, when using generated depth map for view synthesis.