UK-based video encoding start-up V-Nova has unveiled a proprietary compression algorithm called Perseus, which it claims uses a set of technologies that are completely new to the industry, and is a competitor to industry standard MPEG’s next-generation codec HEVC. While little has been disclosed about the technical inner workings and processes behind its compression V-Nova claims that Perseus is more efficient and scales better than HEVC.
V-Nova’s first claim is that Perseus is entirely novel in process and structure. In reality this means that it doesn’t infringe upon any currently enforceable patents, most notably for existing compression standards such as MPEG-2, h.264, HEVC, VC-1 or JPEG-2000; that Perseus is protected by at least one of V-Nova’s 30 claimed patents; and that there is no example of prior-art (previous patent or implemented example) of the patent before publication of V-Nova’s filings.
V-Nova’s second claim is that it has developed a very efficient and highly scalable compression method that is structurally capable of more than HEVC. This takes two forms: efficacy of compression, and structure of the processes allowing it to be scalable. During the NAB Show 2015 in Las Vegas, Perseus was demonstrated as a wrapper on h.264 and in its pure, and allegedly entirely novel, full implementation. The full implementation streamed a live 4Kp60 signal at contribution quality at 300Mbps with 3 frames of latency, a wrapper version on h.264 at distribution quality at 12Mbps which scaled to a 1Mbps HD version on a mobile phone (both with around 30 seconds latency), and an SD signal streamed at 158Kbps (including transport stream). This last version essentially allows video distribution over 2.5G mobile networks.
V-Nova claims to have created such highly efficient compression with a novel process by starting from scratch in working out how compression needs to work given the computational resources available on modern silicon. Major standards, and particularly the MPEG family (MPEG-2, h.264 and HEVC), have been built upon the same framework of linear, serialised processing that was available to the original MPEG-2 implementation in the 1990s. V-Nova is claiming that it has been able to formalise structures and processes which are only possible with multi-core silicon and therefore have not been fully appreciated or applied in any previous compression standard, making for a more efficient compression
The logic of the need for new compression is largely undisputed. While MPEG-style compression has proven very effective, it is likely that there are better processes than macro-blocking and DCT by which compression can be achieved at scale, and which would not have been possible to put on consumer silicon in the 1990s but are viable on today’s silicon. Therefore it seems very reasonable to assert that V-Nova has done something clever and has identified a longer term issue around existing compression formats. What is less clear is whether it has done so in a way that will be both successful in the market and defensible.
V-Nova’s core claim is that the company has developed a means to compress video that does not infringe existing enforceable patents. The nuance in this claim is that it does not require that V-Nova doesn’t use any patents at all, simply those that are not in current term and are therefore in the public domain and cannot be claimed against. The largest pool of general video compression patents were all filed prior to 1996 in order to be added as essential to the MPEG-2 pool published that year. Both main patent agencies in Europe and the US have general terms of around 20 years. Given that MPEG-2 would constitute prior art on any essential patents filed after the publication of the standard, it would mean that all patents essential to MPEG-2, which captures many of the underlying concepts essential to video compression, have in reality expired in the last 2-3 years. It could be no coincidence that V-Nova has come out with compression technology as many core video processing patents expire and become license-free. V-Nova could even be the first of several companies exploiting the new freedoms allowed by MPEG-2 patent expiration.
Determining whether Perseus is actually novel is virtually impossible without more openness about how it works and is implemented. In reality it can never be more than a claim until an implementation has been tested in court; essentially where an owner of another patent will claim that it is necessary for implementing Perseus. If even one patent that is not owned by V-Nova is proven to be essential to Perseus and cannot be worked around then there would be a second party with rights to a royalty payment for Perseus use. The issue of prior-art is probably less problematic since all that V-Nova needs to determine is that at least one patent is novel and therefore can be licensed. Even if 29 fail, that one would be enough to command a royalty payment in every country covered by that patent where Perseus is implemented or deployed.
It is likely that the market will start by testing if Perseus infringes existing patents, rather than trying to disprove the novelty of V-Nova’s of claims; a notoriously esoteric process since it is almost by definition that unpatented prior-art is not well known or easily discoverable. The risk associated with this is not only to V-Nova, but also to any company willing to implement Perseus since infringed patents can apply for retrospective licensing. There are many cases where significant fines are leveraged against companies for implementing someone else’s patent which they were not even aware of. Examples such as Tivo’s successful case against Dish for DVR implementation or Rovi’s ongoing and historic cases protecting its grid guide patents showcase how high the risk is for any operator or consumer device maker that would be liable for patent payment were they to implement V-Nova. One of the main benefits of patent pools like MPEG-LA, Sisvel or more recent poolings like HEVC Advance typically mean that infringement risk is limited by the process of pooling essential patents.
Notwithstanding the product risks associated with patents, the question of market benefit will determine much of the opportunity and success for V-Nova. The simple fact is that HEVC as shown by the industry is not yet fully optimised. There is going to be a process of optimisation which will take several years and will probably result in the kind of efficiency gains seen by Perseus. On the other side, Perseus is likely not optimised either. How far the two processes can go in creating efficiencies is something only time will tell, but it needs a market of vendors implementing onto devices to find out, and HEVC has the investment right now.
Perseus was showcased in two versions: a fully proprietary stack and a wrapper on h.264 with indications that it could become a wrapper on HEVC as well. If Perseus deploys its full proprietary version into the consumer market then it’s very likely its patent claims will be tested, either behind closed doors or in court. Widespread consumer device licenses typically take around $0.2-$2 per device and can transact amounts counted in the low hundreds of millions of dollars easily, especially if deployed on scale devices like smartphones, tablets and TVs. By contrast to completely new standards wrappers are widespread and provide efficiencies, normally on core h.264 decompression on hardware in consumer devices, while differentiating and specialising for needs with software. Proprietary wrappers like DivX do this, and open standards like Xvid or mkv do too, often allowing for many different core compression versions such as including license-free codecs such as Dirac and VP9. Perseus might end up as just the latest wrapper on the newest main standard, HEVC. The general expectation with wrappers is that they will offer 30-50% gains on the core standard, making V-Nova’s claim essentially par for the course. And while wrappers do offer efficiencies, the lower barrier to entry has meant that the industry isn’t as willing to invest in wrappers or to deploy them as widely. The most widespread licensed h.264 wrapper is probably DivX, which seems to run its programme at $10-20M per year.
It is worthy of consideration that as the ecosystem grows wider, compatibility within that ecosystem grows more important. There are many proprietary compressions in video capture and post-production, fewer in contribution and virtually none make it to consumer devices. Perhaps Perseus is better suited to mobile and vanilla HEVC to more widespread distribution. Perhaps Perseus makes sense in contribution encoding but doesn’t scale to consumer devices. Likewise, perhaps Perseus makes enough sense to win over a market-defining deployment such as Apple, Netflix or YouTube, but perhaps not. Perhaps the license will be too high, perhaps the efficiencies too low. However, one thing that V-Nova makes abundantly clear is that MPEG-styled compression is not fit for purpose longer term and the industry will need to rethink how compression works sooner rather than later. V-Nova might be a significant player in that process, even if it doesn’t end up owning the whole thing.