360° Video Coding using Adaptive Tile Partitioning

Encoding 360° video with ultra high definition requires high bit rate to guarantee either immersive experiences and acceptable QoE in video delivery services, or high performance in machine vision applications. However, since in general the full Field-of-View (FoV), i.e., 360°, is not required at once, a great deal of bandwidth can be saved by allowing partial decoding of limited size FoVs. The conventional approach to accomplish such goal has been to encode several fixed-size independent tiles of each video frame. In this work a novel tile-based video coding scheme is proposed to achieve further reduction on the average maximum bit rate required for partial delivery of 360° video. The proposed method dynamically adapts the tile sizes to the omnidirectional video content in order to obtain uniform bit rate in each one. Therefore, more complex image regions, i.e., those requiring higher bit rates are partitioned in smaller tiles, allowing finer bit rate granularity when FoVs with higher spatiotemporal complexity are selected for transmission and decoding. The simulation results show that the proposed adaptive tile-based coding mechanism outperforms the conventional fixed-size tilling methods, achieving an average of 4.78% and 16.88% bit rate reduction for FoVs of 90° × 90° and 45° × 45°, respectively.