ABSTRACT
To achieve improved quality of service and efficiencies with higher resolution, high dynamic range (HDR), and wide color gamut (WCG) in online media delivery, there is a need for advanced video compression standards.
Alliance for Open Media (AOM), a joint development foundation, is targeting AOMedia Video 1 (AV1) as a royalty-free video coding format that is likely to be finalised before the end of 2017. Built on top of Google’s VP9 codec, AV1 has brought in new coding tools from other open source royalty-free codecs such as Google’s VP10, Cisco’s Thor, and Mozilla/Xiph.org’s Daala.
Multiple new royalty-free coding tools have been contributed by members of AOM. Tools that have been legally vetted and technically provide good coding gain to decoding complexity trade-off have been consolidated as default tools, while many other tools are still available as experimental tools.
In this paper, we provide an analysis of the coding gains offered by the default and experimental tools and the corresponding decoding complexity increase for the over-the-top (OTT) adaptive bit-rate streaming delivery use-cases. Also, the coding gains are compared against the compression performance of x264, x265, and libVP9 open-source codecs to highlight the potential bit savings possible with AV1 when migrating from these previous generation options using multiple video quality metrics such as PSNR, SSIM, and Netflix’s VMAF metric.
These results indicate that AV1 offers competitive compression performance over H.265 without significantly increasing the decoding complexity. Though the encoding complexity at this stage is several factors higher, multiple encoding presets exist that trade encoding complexity for reduced compression gain.
INTRODUCTION
The ability to watch personalised content anytime and anywhere at ever increasing resolution, framerate, and quality has resulted in online media delivery gaining momentum over traditional broadcast delivery.
In spite of the progressively increasing broadband bandwidths to subscribers, the cost of unicast delivery to the online media delivery service providers of a large array of titles in ultra-highdefinition (UHD) HDR-WCG necessitate higher video compression. While the H.265/HEVC standard offers in excess of 30% bit savings over the well-entrenched H.264/AVC standard, its adoption is mired in unclear royalty payments to organisations with essential patents.
In September 2015, AOM embarked on developing a royalty-free standard that achieves a compression efficiency that far exceeds that of H.265/HEVC. The codebase used as the starting point for developing the first version of the standard was Google’s VP10 that had already extended Google’s libvpx codec with new coding tools. Aspects from other royalty-free codecs such as Thor and Daala were integrated into this codebase. In addition, many new tools have been experimentally integrated into it.
The tools are legally studied to ensure that they are good candidates for a royalty-free standard. As of April 2017, several of these tools that have offered good compression gains without disproportionate increase in decoding complexity have been enabled by default, while the remaining experimental tools are being refined.
This paper presents an analysis of the compression efficiency offered by the different tools in the default set and in the experimental set. These are compared against the popular open-source implementations of H.264, H.265, and VP9, namely, x264, x265, and libvpx. Though some comparison in this regard has been done in, since newer tools have been consolidated recently, and also conditions like intra-period are different, this analysis is expected to provide different results compared to that.
No comments yet