Towards Much Better SVT-AV1 Quality-Cycles Tradeoffs for VOD Applications

SPIE Optics + Photonics (SPIE)


Software video encoders that have been developed based on the AVC, HEVC, VP9, and AV1 video coding standards have provided improved compression efficiency but at the cost of large increases in encoding complexity. As a result, there is currently no software video encoder that provides competitive quality-cycles tradeoffs extending from the AV1 high-quality range to the AVC low-complexity range. This paper describes methods to further improve the SVT-AV1 overall quality-cycles tradeoffs for high-latency VOD applications.

The paper first presents the latest SVT-AV1 encoder evaluation results generated using the conventional convex-hull-based evaluation approach. To better account for all the content being considered in the evaluation, the combined convex hull approach is then introduced and corresponding results outlined. A more practical version of the latter approach, referred to as the restricted discrete approach, is then presented. In this approach, only few discrete quality levels that are of relevance to typical adaptive streaming applications are considered. This restriction makes the evaluation data more relevant to real adaptive streaming applications and reduces the storage requirements for the encodings generated at the identified optimal encoder settings. A fast version of the latter approach is then presented, where the generation of the optimal encoder parameters is performed using a fast encoder, and the final encodings needed in the evaluation are generated using the desired encoder based on the identified optimal encoder parameters. These optimizations result in a complexity reduction by up to a factor of 10 for at most 1.5% loss in BD-rate.

Featured Publications