8K VVC Encode-Decode Demo
In a demo performed at 3IT Innovation Center for Immersive Imaging Technologies, which is part of Fraunhofer HHI in Germany, Spin Digital, in cooperation with Fraunhofer HHI, demonstrated that 8K content could be efficiently encoded and decoded using the Versatile Video Codec (VVC) from MPEG/ITU. VVC is the successor codec for HEVC and should eventually achieve a 50% bit rate reduction vs. HEVC for similar quality. VVC can support more than just 8K content. It is designed to encode 360-degree content and offers spatial scalability (i.e., adding layers to build up high-resolution frames).
For the demo, some 8K content from Berlin was first captured by HHI and Interdigital. Ben Bross from HHI provided some additional details on the content capture and process.
| Camera: | RED DSMC2 |
| Sensor: | Helium 8K S35 35.4 Megapixel CMOS Sensor 29.90 mm x 15.77 mm (Diagonal: 33.80 mm) |
| Lense: | Fujinon XK6x20 (20-120mm) |
| Recording resolution: | 7680×4320 @ 60fps |
| Recording format: | R3D raw (REDCODE 12:1) with RED Log3G10 log curve and REDWideGamutRGB color space |
From RED raw format, InterDigital generated the HDR10 version using an internal tone mapping process with the output being:
- YUV 4:2:0 10-bit
- BT.2020 color primaries
- PQ10 transfer characteristics
The finished content was then encoded by two different encoders and configured to produce similar quality and operate at a comparable encoding speed. MulticoreWare’s x265 open-source encoder performed the HEVC encoding, with a very slow preset and a 50 Mbps bit rate. The VVenC encoder is an HHI-optimized encoder based on the VTM reference software. It adds features like multi-threading, and subjective optimization based on the xPSNR model also developed at HHI. A medium preset was selected for the VVC encode with a 25 Mbps bit rate.
The main specifications of the demonstration are summarized in the table below.
| Properties | 8K demo specs |
| Video resolution and frame rate | 7680x4320px – 60 Hz |
| Chroma format and bit depth | 4:2:0 10-bit |
| HDR transfer function | PQ (SMPTE ST2084) |
| Color gamut | BT.2020 (BT.2100) |
| VVC encoding | VVenC v1.0 – medium preset: 25 Mbit/s |
| HEVC encoding | x265 v3.5 – very slow preset: 50 Mbit/s |
| HEVC and VVC decoding and player | Spin Player HEVC and VVC |
| CPU for decoding | Intel Xeon Gold 6252 (24 cores) |
| GPU for rendering and HDMI 2.1 interface | Nvidia RTX 3060 |
| 8K TV | Sony Z8H 8K HDR (XBR-85Z8H) |
In most encoding/decoding demos, the decoded content is formatted as a YUV signal, so a player can then format this raw video into a form the TV or display can use. Spin Digital has a new 8K PC-based decoder/player (see an excellent overview of this product and the needs of 8K players here). The Spin Digital decoder works in real-time to decode the VVC bitstream, which is then rendered into 8K/60p content in HDR by an internal media player. The output is then sent to the 8K TV (Sony) over a single HDMI 2.1 cable.
The VVC decode requires 2-3X more compute than HEVC, but this is easily obtainable with available CPU and/or GPU hardware. Spin Digital’s media player leverages a PC with a 24-core CPU for 8K/60p decoding and playback. But their player can also support 8K/120p playback, as long as the CPU is upgraded to 64-cores, to maintain real-time decoding and playback.
An open-source VVC decoder (VVdeC) is also available from HHI online (www.hhi.fraunhofer.de/vvc). It does not perform entirely as well as the Spin Digital one and is suitable for HD or UHD content decoding. HHI is also working on a player that can take the decoded content, package it into an mp4 container, and process it to streaming apps like HLS or DASH.
Spin Digital evaluated both encodes and determined that the quality of both versions was quite similar, with the VVC encode requiring half the data rate. VVC produces better quality in some sequences, such as sky or water textures. VVCenC also showed superior subjective quality compared to the reference VVC encoder.
You can find a video describing the demo HERE.
