Adaptive streaming of interactive free viewpoint videos to heterogeneous clients

Recent advances in video capturing and rendering technologies have paved the way for new video streaming applications. Freeviewpoint video (FVV) streaming is one such application where users are able to interact with the scene by navigating to different viewpoints. Free-viewpoint videos are composed of multiple streams representing the captured scene and its geometry from different vantage points. Rendering non-captured views at the client requires transmitting multiple views with associated depth map streams, thereby increasing the network traffic requirements for such systems. Adding to the complexity of these systems is the fact that different component streams contribute differently to the quality of the final rendered view. In this paper, we present a free-viewpoint video streaming system based on HTTP adaptive streaming and the multi-view-plus-depth (MVD) representation. We propose a novel quality-aware rate adaptation method for FVV streaming based on a virtual view distortion model. This view distortion model represents the relation between the distortion of the texture and depth components of reference views and a target virtual view and enables the streaming client to find the best set of representations to request from the server. We have implemented the proposed rate adaptation method in a prototype FVV DASH-based streaming system and performed objective and subjective evaluation experiments. Our experimental results show that the proposed FVV streaming rate adaptation method improves the user's quality-of-experience and increases the visual quality of rendered virtual views by up to 4 dB for some video sequences. Moreover, users have rated the quality of videos streamed using our proposed method higher than videos streamed using other rate adaptation methods in the literature.