Bitrate Reduction for Omnidirectional Video Streaming: Comparing Variable Quantization Parameter and Variable Resolution Approaches

摘要

Immersive and Extended Reality applications are getting increasingly popular due to the recent improvements and wide availability of Head Mounted Displays in the mass market. Among the many new immersive applications, Omnidirectional Video Streaming (OVS) - or 360°Video Streaming - is attracting the attention of both the industry and the research community. The increased immersivity of 360° videos comes at the cost of larger bandwidth requirements compared to classic 2D videos. To tackle the issue of reducing bandwidth requirements, tiling is a viable technique that allows to encode the portions of the 360° video most likely to fall outside of the users' viewport at lower quality using a higher quantization parameter. Tiling requires new encoders to be used which however do not have available hardware decoders in mobile devices yet. The variable resolution approach instead shrinks areas not falling in the region of interest to decrease the overall resolution and thus allowing bitrate reduction with any codec. This paper quantitatively compares the two approaches to find the trade-offs between achievable bitrate reduction and visual quality measured using the VMAF visual quality metric.