Guest Post: HEVC Compression of Video Explained

During episode 2665, we mentioned the BBC research labs confirmation of HEVC compression. During the discussion Scott guessed correctly that Andy Beach would be able to explain this more for us. And Andy obliged. Here’s his response in full.

Like summoning a genie you say “bitrate” or “compression” three times and i appear!

I was literally already jotting down notes as I listened. here’s my thoughts:

The savings side viewed by the BBC team arent astounding – they are actually what was expected out of the codec – even taking the perceptive testing the did. What I am happy about is how quickly we’re getting to those numbers and seeing them in public press.

Let’s take a quick hop back – even back when H.264 was announced as an idea, the goal was to improve upon the current video defacto codec (at the time MPEG-2 part 2 i think) by 50% savings of bits to perceived quality. That was back in 1999 or 98. The spec was completed in 2003 and it was another two years before we really saw people start to take an interest (and really another 2 years before we started seeing real adoption). H.265 on the other hand was first ratified in 2013 (and version 2 which is what most would consider the viable useful version of the codec was only ratified in January of 2015) and here we are right at the start of 2016 and not only have we been seieng demos at shows like NAB for a year or two, but we’re already seeing early adopters like BBC with confirmation of what we expect to see from the codec.

I’m hopeful that we start seeing some software based (encode) support in the market this year (apps like VLC already include H.265 support) and next year a wider announcement of Hardware support decode support. It is the HW decode support that will become a blocker to adoption as it will be required for things like mobile handsets and tablets to take avantage of H.265 and the only really way for someone to get the true value out of encodes targeted to big 4K screens (Because of the nature of video compression, highly aggressive encodes are very resource intensive to decode – if all decode is SW based, its hard to optimize the content down to those very low bit rates we want – adding HW decode support enables improved battery life and performance on mobile and allows bigger screens to really crank those rates down while still keeping a decent quality).

But Andy, does this mean my H.264 archive is useless? Not at all – we’ll continue to see it used and supported for a number of years. Too many devices are out there that already support it and at this point it does what it does really well. Like new TV adoption, we’ll see H.265 support creep out slowly through a variety of expected names (think olympics, superbowl, netflix, etc). Effectively, anyone really wanting to do streaming 4K content will be an earliy H.265 adopter. As they teach us all lessons for how to do large operations at scale with this codec, we’ll see others migrate over. If I were to put on my guessing hat (aka analyst) I’d say between now and late 2017 is early adopter world. Then we’ll see wide expanding support from early 2018 to 2021 or so and by then the market should be well saturated and we’ll be discussing its successor.

What else? There are two obstacles that stand in H.265s way – well maybe three if you count itself and its royalities as an problem to deal with. If the numbers are too high, neither content producers or technology companies will be able to make it a viable solution.

The other major technologies in the works though are Google and their VP9 codec and the AOMedia Alliance. If you talk to Colleen Henry, she’ll tell you VP9 is already the greatest thing int he world and you’d be an idiot for using anything else. The test results I’ve seen put VP9 as a ver close second to H.265 in most cases (and maybe in front of it in a narrow few) but the reality is studios just have a hard time trusting Google and I’d be surprised if they’d willingly adopt it as the “go to” codec. also what DRM capabilities VP9 will support have been unclear and all broadcasters and studios will require (demand) the ability to protect their content or control the licensing and viewing. the AOMedia is a group of big technology companies (amazon, microsoft, intel and the like) who have all loosely gotten together and are working on a spec that would be similar to H.265 but royalty free. From what I’ve seen, it does look like they are potentially using VP9 as the starting framework here, but its unclear how much the eventual product would look like the current one. It’s also fairly early days and honestly this seems to me like a gambit (read as threat) to keep the MPEG-LA (licensing group of MPEG group) at bay when negotiating (meaning “charge us that and we’ll just build it ourselves!”)

Well, that was super long winded, so let me try to summarize a podcast friendly sound bite:

  • The numbers were good and what we would have expected.
  • Its great to see big companies like BBC already going down the H.265 path and it Indiates it may see quicker adoption than its predescessor.
  • Hardware decode support will continue to limit wider deployment and use, but expect to see more and more announcements for it in the next 18 months.
  • Just because H.265 is the new hotness, don’t expect H.264 to dissapear overnight – it’ll be around for a good long time.
  • Keep your eye on Google/VP9 and AOMedia/whatever they come up with as potential competitors to H.265

Did I miss anything? Feel free to ask questions, as you know, i love this stuff!