Adaptive Multi-Resolution Encoding for ABR Streaming

Abstract

Adaptive bitrate (ABR) streaming implies encoding of the same content in multiple spatial resolutions. This work presents a novel adaptive scheme which significantly improves the computational efficiency of multi-resolution encoding. Video content is first encoded at a lower resolution, and encoder decisions, such as motion vectors and quadtree block structure, are then extrapolated and used in higher resolution encodes. Low-resolution encoder decisions are further refined using static and machine-learning based dynamic techniques. The scheme was validated using the open source x265 High Efficiency Video Coding (HEVC) encoder in capped variable bitrate (cVBR) mode, typical for modern adaptive streaming. The proposed adaptive multi-resolution scheme achieves speed-ups of up to 250%, compared to the conventional single-resolution approach. Furthermore, this speed-up is achieved without substantial losses in coding efficiency.