Interactive multiview video system with low decoding complexity
Research in multimedia is always investigating new ways of improving the immersive experience of the users. One current solution consists in designing systems which offer a high level of interactivity, such as multiview content navigation where the point of view can be changed while watching at a video sequence (e.g., free viewpoint television, gaming, etc.). The coding algorithm designed for the transmission of such media streams must be adapted to these novel decoder needs. However, video plus depth data transmission is usually performed by considering the information flows as two sequences encoded with MVC schemes. Whereas it achieves good compression performance, this coding approach is not appropriate for interactive applications since the decoding of a frame often requires the prior transmission and decoding of several reference frames. Moreover, the techniques recently developed to improve interactivity are generally implemented at the decoder, whose computational complexity requirements are augmented. In this paper, we propose a novel coding scheme for video plus depth sequences that is adapted to user navigation; contrarily to several common approaches, the additional complexity is added on the encoder side so that the decoder stays simple. We further propose to limit the additional bandwidth imposed by interactivity requirements by designing a rate allocation algorithm that builds on a model of the user behavior. A first version of our novel coding architecture is evaluated in terms of rate-distortion performance, where it is shown to offer a high interactivity at a reasonable bandwidth cost.