A couple of weeks ago we reported on the launch of V-Nova, the new compression company that came out of a four year stealth period with a new codec called PERSEUS that it claims can deliver UHD in HD bit rates, HD in SD bit rates and SD in sub-audio bit rates (read original story here). This is a proprietary codec that will be licensed openly and used by V-Nova itself for its own compression products. The company gave no indication that it was going to put PERSEUS through a standardization process. Sky Italia already uses the codec for video contribution links between studios in Milan and Rome within an existing standards-based workflow and the Pay TV operator hinted that it was interested in using the codec for its DTH distribution network, too.
V-Nova shocked industry journalists by revealing that it has been operating since 2011, keeping its technology away from public gaze. The V-Nova launch was also a surprise to the wider broadcast market so, a few weeks later, what has been the reaction to the launch? We already reported on how Broadcom is working with the PERSEUS codec, what Elemental Technologies thinks about its arrival and how Wyplay is working on several implementation projects with V-Nova.
Keith Wymbs, Chief Marketing Officer for Elemental Technologies, sums up what others have been saying to us privately. â€œClaims of proprietary codec gains are very common in the video industry. Particularly in the Pay TV segment, it is difficult for proprietary solutions vendors to build out an ecosystem that is robust enough to gain widespread traction.â€
People who have commented about the new codec to Videonet, whether for publication or off-the-record, believe there are three big challenges for the uptake of this new codec and they are the same as would apply to any other. First, the television industry needs standards for the distribution market, even if you might be able to build a closed end-to-end ecosystem for contribution. You cannot build the critical mass of support needed without an interoperable ecosystem.
Second, the codec needs to be independently and vigorously tested in the way that HEVC and Googleâ€™s VP9 have been. Companies have made big claims in the past, which have been found to be limited to very particular video applications. Third, standardization provides transparency about who owns intellectual property that is used in a compression solution and therefore relative safety (against patent infringement claims) to those deploying it.
Iain Richardson is CEO of Vcodex Associates, a consultancy that provides impartial advice on video compression and has acted as expert witnesses in multiple patent cases involving H.264, video compression and image processing technology, and provided source code analysis for litigation. He is also CEO of Beamshare. He picks up on the intellectual property concerns, emphasizing that he is outlining the general challenges to a proprietary solution, having had no opportunity to assess PERSEUS specifically.
â€œPeople are very, very sensitive now to the fact that if they are offered a â€˜black boxâ€™ codec, they donâ€™t know what technologies are in it and who might have patent rights,â€ he observes. â€œSomeone could tell you they built everything themselves but there are thousands of patents on video coding. It is very difficult to build something without treading on someoneâ€™s toes somewhere. With a standard you have gone through the whole process of achieving IP visibility. It is not plain sailing, even with standards, but everyone looks to declare their IP interest.â€
Boris Felts, VP Product & Solutions at Envivio, the multiscreen encoding and transcoding vendor, declares himself very sceptical that any proprietary codec can succeed in the television distribution market, even if it could find a home for contribution. And it takes a lot to compete with an MPEG standard. He points to codecs that resulted from single initiatives, such as VC-1, VP8, VP9 or AVS in China that, despite their openness and being supported by a relatively wide ecosystem (in terms of devices, content or geography) do not have the same device footprint and industry support as codecs that went through MPEG standardization. They usually served a pretty specific application or business case.
â€œFor contribution and video applications like security, proprietary has potential,â€ Felts confirms. â€œWe have seen proprietary codecs for contribution in an end-to-end encoder/decoder set-up and as long as the content partners are okay with a closed system then that is perfectly feasible. But I am less convinced about a non-standards approach for mainstream distribution like satellite or terrestrial where there are a whole lot more devices you have to interact with.â€
Richardson argues that the real need for bandwidth savings is in the distribution, rather than contribution networks. And in the link from headend-to-home, he says, â€œIt is unusual these days for both ends [encoding and decoding] to be controlled by the same player. You need wide support for the codec at both ends [of the delivery chain].â€
This still applies even if you ignore broadcast networks for one moment and focus on OTT video, Richardson believes. He says that connected CE devices use hardware support in order to improve performance and avoid draining batteries when decoding video. â€œThe problem is that devices have hardware acceleration for the incumbent or dominant standard and today that is H.264. You have quite a challenge if you want to get a new black box system on them.â€
He points to the Google experience with VP8 (the companyâ€™s equivalent to H.264) and more recently VP9 (designed as an HEVC equivalent) as an important case study. â€œGoogle has worked extremely hard over the last 4-5 years to get VP8 support out there. They have released reference designs for chipmakers and it is open source. You can get the software free if you are a developer.
â€œWhen something like VP8 or HEVC comes along, over a period of years the codecs get built into hardware and software components â€“ into chipsets and operating systems and gradually support grows until you reach a critical mass. HEVC is the codec that is moving that way.â€
There is one obvious difference between VP8 and VP9 and what V-Nova is claiming for PERSEUS, however, and that is the level of performance gain. Richardson reckons the Google codecs are considered comparable to their MPEG counterparts in terms of performance, based on various independent studies and comparison points. PERSEUS is said to provide efficiency gains of around 300% compared to MPEG standard codecs.
Richardson admits â€œIf you really can provide more than twice the performance [of what exists today], that would be a strong incentive for people to pay attention.â€
And that brings us to the performance claims. At its launch, V-Nova demonstrated UHD video that had been encoded using PERSEUS being distributed at 8Mbps compared to 21Mbps with HEVC encoded content, for what it says was comparable video quality based on key performance indicators. It said that contribution feeds that would require 1Gbps of bandwidth using JPEG 2000 are being achieved with 300Mbps and SD video is being delivered at 300Kbps, making it possible to provide mobile television over 2G networks.
The people we spoke to made it clear that the market needs independent testing to confirm the claims. Boris Felts at Envivio reckons there is a start-up company every couple of years saying they have improved things to disruptive levels. â€œIf they make claims for 50% efficiency improvements everyone looks at them, and even that requires that they would have done something really special and â€˜cracked the codeâ€™,â€ he comments.
â€œWe need objective [test] results,â€ Felts declares. â€œIt sounds like the PERSEUS codec is scalable and so can have multiple resolutions and bit rates in one stream. Previous attempts to create a scalable codec have mostly focused on video conferencing or security video but not mainstream broadcast because they are not as efficient for single rate resolution coding.â€
Iain Richardson hopes that V-Nova have made a breakthrough, confirming what the new compression company said at its launch, that VP8 and VP9, and H.264 and HEVC are using the same basic processing model but optimizing it further with the new generation of compression. â€œI would love to see someone come up with a radical advance.â€
So again, it comes down to proving the performance claims. Richardson recalls: â€œI have come across many situations in the past where it is possible to show really good performance if you control the conditions for the test, including the content used and the bit rate target point, but it proves much harder to replicate that gain across a broader range of video requirements.
â€œYou need to allow someone or a group of people to test the codec independently and to really stress test it with a range of sequences and a range of video material, then use objective quality measures like signal-to-noise ratio and subjective testing with viewers to assess different reference points. You need the kind of testing that HEVC and VP9 have been exposed to.â€
Ericsson, an encoding leader for contribution and distribution, sums up much of what has been said. â€œWe are unable to comment on the performance of the PERSEUS codec as we have not seen any independent testing results. But Ericsson is firmly committed to developing and advancing interoperable standards for the industry in order to promote and enable success for our customers.â€