Compress2X Offers New Approach to 8K Encoding
(Compress2X has just renamed itself – it was previously known as Wauwtec).
The world of video codecs is a fascinating one if you are a student of technology development. From the early days of MPEG 1, thirty years ago, mainstream codecs from the MPEG group have developed dramatically. MPEG-1 had significant quality limitations and supported only 1.5Mbps (although the audio layer is the very widely supported MP3 format).
MPEG-2 was adopted very widely (including for DVD content) and added support for High Definition video. MPEG-4/H.264/AVC (3 was merged back into MPEG-2) was next and used for Blu-ray and broadcast. Then came MPEG-H Part 2/H.265/HEVC, followed by MPEG-I Part 3/H.266/VVC.
Each Generation Brings ‘More of the Same’
Each generation, broadly, did ‘more of the same’ with researchers developing and enhancing the technology techniques and processes used to exploit the increased power of semiconductors to encode and decode the increased complexity. We dug into the complexity of VVC in a discussion with Spin Digital which is one of the first to be able to encode in VVC for live broadcast.
Each of these generations has typically allowed a reduction of 50% in bitrate for a particular level of visual quality. As bitrates come down, with better codecs, higher quality video can be provided at a similar transmission cost, although compressing and decompressing can take more power.
Codecs Double, but Pixels Quadruple
One of the challenges for higher quality video is that while codecs typically double in performance with each generation, moving from FullHD to UltraHD (4K) and then to 8K each involves a quadrupling of the number of pixels (so from FullHD to 8K is an increase by 16 – 4 X 4). Add higher frame rates and greater bit depth (for HDR and wide color gamut) and keeping the bitrate down can be a challenge as each feature adds to the number of bits to be processed. As we reported before, NHK’s early tests of 8K needed 85 Mbps and with improved codecs, that has reduced to around 40Mbps, but that’s still a lot compared to the 10-15Mbps that satellite operators and streamers like for premium content.
So when we hear about a new approach to codecs, we tend to pay attention, especially when they start from a different place to the standard ‘incremental’ process used for the main codecs. Back at the end of last year, we heard of a codec company, Compress2X based in Eindhoven in the Netherlands, but with offices in several countries around the world, that claimed a dramatic reduction in required bitrates compared to the MPEG approach. The firm said that it could transmit good quality 8K at 12.5Mbps.
The last time the author heard such a claim was more than ten years ago from a technology start-up called V-Nova. That company said that it thought that a fresh approach was needed rather than the incremental development of ‘more of the same’. It made dramatic claims, but over time, its technology was accepted and became standardized as MPEG-5 Part 2 Low Complexity Enhancement Video Coding (LCEVC). LCEVC is a codec-agnostic enhancement process to create a layer that can be combined with other video codecs to improve their coding and processing efficiency. It also maintains compatibility with the device ecosystem of the enhanced codec.
Compress2X Keeps its Technology Secrets Close
So, sometimes ‘left field’ approaches to codec development have paid off.
Compress2X was cautious about revealing the technical details of its technology. The firm did tell us that its different approach is not based on the kind of frame-to-frame prediction that other codecs use, but is built on frame by frame calculation using a complex algorithm that has taken ‘twenty years of work’, the firm told us. Like LCEVC, the firm said that its technology can be used in conjunction with existing codecs and streaming and broadcast protocols. That’s important as it means that the streams can be decoded by existing standardised decoding chips.
At the moment, the firm’s compression system works only on Apple Mac PCs, although it is working on supporting other platforms. It has seen interest from major broadcast codec hardware suppliers and is also working with AWS on virtual compression applications in the Cloud. The firm was keen to highlight to us that both local or cloud-based encoding is possible.
Good Quality Compression
Now, anybody can claim great compression factors of a codec, if they are prepared to sacrifice the image quality. To demonstrate its technology, Compress2X told us that it had encoded a sample video, that the 8KAssociation had provided, at different levels of compression. The compressed clips were evaluated by Ssimwave, these days owned by IMAX. Ssimwave is a specialist in putting numbers onto the thorny question of how good video looks.
We were given access to the report and to confirm what it meant, we spoke to Hoyat Yeganeh, Principal Video Architect and Research Lead at Ssimwave who has ten years experience of video quality metrics. He helped us to interpret the report.
Compress2X submitted a number of files – a proxy source provided by 8K Association and nine versions, encoded at bit rates from 78 Mbps to 12 Mbps. Ssimwave’s system of metrics has to be matched with an output device – after all, content viewed on a smartphone will look very different than content shown on a big TV. The largest and highest quality target device that Ssimwave had was an UltraHD OLED TV from LG (67C9PUS), so the measurement was based on viewing on that device.
The scores for the Encoder Performance Score (EPS – one of the key Ssimwave metrics that measures the difference between the source and the content being tested – irrespective of the original quality of the content) showed high scores of from 86.1 (12 Mbps) to 98.8 (48 Mbps). Yeganeh confirmed to us that anything above 80 is considered ‘excellent’ by his firm. (Strangely, the files at 56 Mbps and 78 Mbps were not quite as highly ranked as the 48 Mbps sample).
The Ssimwave system allows individual frames to be viewed and analyzed with the error data revealed and visual inspection confirmed the scores, Yeganeh told us.
Banding An Issue
Ssimwave has a separate metric for banding (although it is developing a way to integrate that with its EPS score) and here the result was not so positive, with banding of some frames in the range of 70-88. The 12 Mbps band, in particular contained ‘very high levels’ of banding. A higher score means more banding, and Yeganeh told us that the firm thinks that anything above 60 ‘is an issue’. Again, inspection of the worst offending frames confirmed the banding. The writer also had access to the files and saw visible banding on the 12Mbps stream that was not visible in the highest quality streams.
So in summary, Compress2X has an interesting approach to the encoding of compressed video at quite low bit rates. The fact that a high EPS could be achieved at 12 Mbps (albeit with some banding) is indeed impressive, although the target device was an UltraHD set rather than an 8K set in the Ssimwave tests.
The Compress2X encoder has been developed based on the Apple Mac platform and the firm already has a global partnership with AWS and can support ‘Virtual macs’ as well as local systems. It is developing a user interface for the AWS platform. In terms of the speed of processing, at FullHD resolution on a Mac Pro, it’s real time, although slower for higher resolutions.
Compress2X will be at IBC, as will Ssimwave as part of the IMAX presence on Stand 8.D45
You can also reach them at firstname.lastname@example.org