The mysterious V-Nova codec start up, which came on the scene in early 2015, has acquired one of the other specialists we have been following in compressing video data, Faroudja Enterprises, and in particular its patents.
Both have made bold claims about compressing UHD video by 35% to 50% more than a standard HEVC encoder, and because of how they work, they will be “additive” – in other words the outcome should be a stream that is between 25% and 45% of the bandwidth of an HEVC encode.
Neither company has explained their respective patents in great detail, why would they, but they have both spoken to Faultline Online Reporter in the past, with V-Nova CEO and Co-Founder, Guido Meardi dropping hints about how V-Nova’s Perseus works, and Faroudja’s COO Bill Herz, giving us a more precise run-through of its patents, at its launch.
Essentially Faroudja pre-processes a file or stream before it goes to an encoder. It is trying to reduce the amount of data that has to be encoded. It has two output steams, and sends the second one on ahead to be “re-introduced” to the decoded video stream, inside the player. This stream also goes through the encoder, and then this data is put back where it belongs in a post processing phase in-device.
It works by taking away half the lines in a screen, uses some algorithms to estimate what should or must be there, and transmits only the differences between the forecast and what was actually there.
Its patent talks about separating horizontal high and low frequencies from the video signal using low-pass filtering, and outputting an information-reduced signal. Then the filtered component generates a second supporting signal, which tells the remote devices whether to add frequencies to the output and to what extent.
It also deletes every other line in the vertical direction and a similar filtered signal is sent, also added back, and then again, every other frame is also treated in the same way.
It looks to us like the codec only has to encode half the frames and only about 25% of the data in the between frames, for the weaker of the two signals, but the penalty is the additional separate parallel data stream that needs to be send with extra detail.
This is taking a very non-standard, dynamic encoding process which takes more processing to produce as say another HEVC encode pass, and adding it as a pre-processing stage. After that the data that is extracted, essentially the “fine detail” of the image, is sent separately, both encoded.
Although V-Nova says that the gains from this process are additive to its own Perseus codec, it also uses a higher degree of prediction somewhat akin to a more extreme version of the CABAC mathematical movement projections, which are included in Perseus. But let’s say that they are “mostly” additive for now.
Faroudja managed to compress a 4K 30fps stream using an old H.264 encoder, down to 8 Mbps (compared to the Netflix 4K compression of 15 Mbps). If it was used alongside a V-Nova codec working either in native mode, or in parallel with an HEVC codec, then the potential is there to send 4K in something like 4 Mbps, maybe less.
Both use a level of pre-processing, but both have argued that modern CPUs have plenty of capacity for this, and in V-Nova’s case Meardi suggested that it uses less battery than existing handset decodes when reading the files.
There were no details of the acquisition and given that Faroudja had a really low employee count, we suspect that little or no money has changed hands, but that V-Nova has simply convinced Faroudja that its inventions will come to a great financial fruition if combined with its own. Potential it was simply a share swap deal.
This is almost because straight after the two companies launched in 2015, the market was thrown into turmoil, initially with two separate patent pools for HEVC, and then the launch of a completely different approach to encoder development, one of continuous development and updates, from a royalty free consortium called AOMedia, with backing from Google, Microsoft, Cisco, Netflix, Amazon, Mozilla and others.
HEVC is then caught in a three-way battle between MPEG-LA (Apple, BBC, Fujitsu, Hitachi, Humax, Orange, NTT and more), HEVC Advance (Dolby, GE, Mitsubishi and Philips), and the newest rival body in The Alliance for Open Media (Cisco, Google, Microsoft, Amazon, Intel, Mozilla, and Netflix). V-Nova wants to rely on license fees, in a market which has formed the impression that it no longer needs to pay them.
The entire compression market has since stalled with almost no-one taking a license from either patent pool, and their financial terms being lowered with each passing month to encourage the market. This has overshadowed start-ups in the area, and the only real hope for these companies is that one of the majors backing the AOMedia grouping will acquire them, and include their patents on a royalty free basis to the AOMedia group, which is expected to have a first cut of its codec in March this year.
Codecs have once again become critical to business development, as 360 video and Virtual Reality are both going to require the continued reduction in stream or file sizes, in the near future, and entirely new approaches to encoding a VR stream.
Just about a year ago, Beamr, a compression rival from Israel, acquired its own HEVC codec, by buying Vanguard Video. Beamr was already claiming bit-rate savings of up to 50% for video of the same perceptual quality, meaning that viewers shouldn’t be able to spot the difference despite the bandwidth compression. Its secret sauce was in fact a number of algorithms which eliminated any detail which the human eye would not be able to notice, in a quest to pre-process the stream to a smaller bandwidth.
Companies this size find it hard to navigate the entire encoding eco-system, and this was a way of getting control over an HEVC codec, which previously Beamr could not get sufficient access to.
And all this comes at a time when many of the leading encoder firms – specifically Elemental, Envivio and Thomson, were acquired out of the market to strengthen other portfolios. There is room in VC circles for an innovator who may be able to take their places, as a prospect for funding.
If the codecs of AOMedia reach market and are rejected, then there may be renewed interest in V-Nova and other start-ups in this space, but in the interim, the market is holding its breath to see what it can achieve.
Of all these start-ups, only V-Nova has managed to find firm video specialist customers in the form of Sky Italia, which uses its codec to compress contribution feeds, and it compresses video telemetry internally at Airbus, and is rumored to play some part in the Tata Sky OTT services in India, through an integration with French set top middleware player Wyplay, although that relationship is far from clear. It’s not a huge haul for two years into the market – but better than both Beamr and Faroudja had so far managed.
Backing for V-Nova comes from McKinsey consulting and what it claims are a variety of major backers, who are not comfortable revealing themselves at present. Hence our description of it being mysterious, as we are unsure of precisely who is behind it.