Options
2021
Conference Paper
Titel
Objective Evaluation of the Practical Video Encoders VVenC, x265, and aomenc AV1
Abstract
Practical open-source video encoders implementing different video compression standards often provide various operation points for different trade-offs between compression efficiency and encoding time. This paper presents an evaluation of the trade-offs for the freely available open-source VVenC encoder conforming to the open and public Versatile Video Coding (VVC) standard together with x265 conforming to the open and public High Efficiency Video Coding (HEVC) standard and aomenc conforming to the AOM Video 1 (AV1) specification developed within an industry consortium. The evaluation includes the HEVC and VVC reference encoder implementations to analyze the tradeoffs provided by the practical encoders when focusing on high-resolution 10-bit consumer applications that require random access. Compared to the HEVC reference implementation HM, VVenC provides averaged objective bit-rate savings using the Peak-Signal-to-Noise-Ratio (PSNR) metric ranging from more than 37% at less than 68% encoding time to more than 9% at less than 5% of the HM encoding time. Sampling the data for an encoding time next to but lower than that of the HM and using HM's outcome as the anchor, VVenC provides 37.6% averaged bit-rate savings, followed by aomenc with 11.5%. In contrast to VVenC that provides the same performance as the VVC reference implementation VTM at a lower encoding time, x265 generates an average of 35.2% bit-rate overhead relative to its reference implementation HM and does not achieve the HM performance even for higher encoding times.
Konferenz