Your browser is not supported. Please update it.

22 September 2022

Weak resistance won’t reduce LCEVC to rubble

Faultline’s surprise revelation at IBC 2022 that opposers (albeit few) to the LCEVC standard still exist was an unfortunate distraction from the work going on inside the MPEG-5 camp during the event – which is a growing force with a now 27-company-strong hivemind.

While we have been privy to numerous LCEVC demonstrations, newer visitors were impressed, we are told, by demos of LCEVC-HEVC at the IBC booth. As well as bandwidth savings, the quality benefits of LCEVC as an enhancement layer to HEVC encoding were even visible at a distance, with a distracted gaze.

So, while last week’s IBC write-up post-IBC was opinion-oriented, let’s attempt to pen a follow-up on LCEVC without deviating too much from the facts.

The first thing to highlight is V-Nova’s announcement at the beginning of the show that it has added the LCEVC Native Development Kit (NDK) to the existing SDK. LCEVC NDK is essentially a toolbox accelerate the adoption of LCEVC among operating systems, browsers, SoC makers and ASIC suppliers.

This is significant on two levels. Firstly, none of these suppliers will have to pay a penny in licensing fees for LCEVC NDK thanks to the LCEVC licensing structure, which accrues royalties from service providers which pay a small and capped fee. Secondly, LCEVC NDK availability marks the first time that implementations can range from full software to silicon.

Back to the battleground now. At IBC, we were pointed to a couple of academic papers by a senior executive at a sizable encoding company to support their argument that LCEVC will not succeed. However, upon closer inspection, the nature of rigorous testing carried out by researchers in such studies makes it easy to cherry pick particular results to fit any agenda.

For the record, Faultline does not have member access to the full paid studies. But we do have access to one from our sister service, Rethink TV, which projects that LCEVC will grow to an attach rate of 30.2%, acting as an enhancement layer for other video codecs – with close ties to AV1 propelling its initial growth.

At the beginning of this forecast, LCEVC is mapped closely to the penetration of AV1. Over the period, this extends to AV2, and eventually extends considerably beyond where Rethink TV expects AV1 and AV2 to finish – at just over 30.2% of the total transcoding market, compared to 19.9%.

A small caveat is that LCEVC is described as the most volatile technology being forecast in the global transcoding report, from a revenue standpoint, and therefore a dollar value has not been calculated.

As for the academic papers, we were informed that the research from The University of Kingston and TU Berlin concluded that LCEVC brings no meaningful mean opinion score improvements for HEVC (x265) against multiple bitrates. However, we have since learned that this was carried out using a 1+ year-old version of the LCEVC SDK. The aging SDK lacked calibration with x265 and therefore to say LCEVC showed no meaningful improvement for HEVC does not account for the bigger picture.

Even so, despite the calibration issue, results showed some modest processing acceleration and quality improvements with HEVC. Some have highlighted that these were lower than normal, which again can be attributed to the lack of calibration with x265, plus high-bitrate operating points.

What is alarming about this is not the subjectiveness of objective research, but how quickly an academic paper can be cast aside as outdated.

This is evidenced in the second paper referenced last week from the SPIE, which analyzed data from 650,000 encodings of 140 input sequences, over many months with Meta and Intel. In addition, the MPEG independent validations were done on a multitude of different sequences and thoroughly cross-checked.

In comparison, the Kingston/TU Berlin report was carried out by just two researchers, on only a handful of sequences.

While we have a duty to report on comments made by senior technology executives (when they are not flagged explicitly as off the record), it can often feel anathema to all the hard work that has been achieved when resistance rears its ugly head, particularly for a movement like LCEVC which has overcome so much adversary.

With success within touching distance, such comments are perhaps unnecessary, but still we must respect that everyone is entitled to their opinion – and we will always try to strike a balanced argument.