A View-Level Rate Distortion Model for Multi-View/3D Video

Multi-view/3D video is currently available in games, entertainment, education, security, and surveillance applications . Since the amount of data in multi-view/3D increases proportionally with the number of cameras, and due to different bandwidth and playback capabilities of receivers, appropriate compression of multi-view/3D video to produce the correct bitrate while maintaining smooth video quality is crucial, a task that is mostly performed by the rate control module of the encoder. There are many existing rate control algorithms for single-view and multi-view video coding considering the specific features or aspects of these videos. In this paper, we introduce a novel view-level rate distortion (RD) model. We use a systematic methodology to derive this RD model by investigating the impact of multi-view/3D video characteristics on the bitrate of a compressed video. Our proposed RD model considers the concepts of intra-view and inter-view disparity as an effective feature of multi-view/3D video to estimate the overall bitrate of each view more accurately. Evaluation results indicate that our proposed view-level RD model outperforms existing linear models by a factor of 3 and can predict the rate of each view with relatively high precision and a low estimation error of 12% on average.