How HEVC could remake Internet video—or not

1/25/2013 1:34 PM EST

VP9 is not the first proprietary codec, nor is it the first to be offered for free. Industry tends to shy away from uncertain licensing regimes for fear of bait and switch or worse, essential patent holders surfacing later. The MPEG model creates a level of comfort that has obviously worked in the past. And in attendance at MPEG are the best minds in compression and reps from various stored program platforms (other than x86 machines) to ensure wide usability and the best efficiencies. It is doubtful that any one enterprise, even one as well staffed as Google, can muster this sort of expertise and broad agenda. And the feature that is mentioned above is an encoder implementation component that any codec can be built with.

1/28/2013 11:19 AM EST

Yes, cable MVPDs still use MPEG2. Many systems are constrained for bandwidth, but the economics of migrating all those set-top boxes to H.264 clearly has not yet been favorable -- like you said, too much hassle for too little gain.

I agree that HEVC will be essential for 4K UHDTV to become mainstream, but I think that even without UHDTV, HEVC will be a huge boon for mobile video. Yes, internet TV in general is hampered by speed issues for all but the fastest broadband connections, but it is far worse for the average mobile viewer.

Even when LTE becomes more ubiquitous and mobile download speeds are less of an issue, carriers are moving toward pay-for-use data plans, which will still give most users reason to pause before watching lengthy videos over a cellular connection. But with HEVC cutting the amount of data in half, that constraint is greatly reduced.

1/28/2013 12:52 PM EST

Understood re cable guys still on MPEG 2.

As I understand it from Setos, its the satellite guys (along with Blu-ray) who helped kick start MPEG 4 Part 10 for HD.

1/28/2013 7:10 PM EST

Good point about wireless broadband in mobile devices. That would benefit from all the compression efficiency possible, and the handheld devices have a short lifespan anyway. So that might help push HEVC along.

But in these applications, the improvement won't be to cut bit rate by half. These systems already use H.264. I doubt any are still doing MPEG-2 compression? Going by the popular guesstimates, I'd expect no more than a 25-30 percent drop maximum, depending as always on the source material. Still significant, though.

1/28/2013 9:29 PM EST

I take claims like that with a grain of salt, Rick.

The whole point of VP8 and VP9 is to avoid having to pay royalties, not to develop a technically better codec. And there's an army of companies developing the MPEG compression schemes.

So I find it very unlikely that improvement over HEVC would be real, long lasting, or across the board. Besides, HEVC encoders will continue to be improved too, as all other generations of encoders have been. H.262 encoders, for example, those used for MPEG-2 compression, are now capable of at least twice the coding efficiency they had when they first came out.

If you look at the techniques used in H.262 vs H.264 vs H.265, you can see that the improvements are quite progressive and I'd say almost predictable. You allow bigger and smaller blocks, you allow more block aspect ratios, you tweak the entropy coding schemes permitted, you allow interframes to be spread further apart, you use filtering techniques to mask blocking artifacts.

The VP series don't take a drastically different approach in achieving compression. They're still discrete cosine transform (or the derivative integer transform) algorithms, so at best, any improvement might be slight. All of these schemes have to deal with the realities of computing power, especially in the encoding devices, so I doubt any real advantage with VP9 (except the royalty issue, perhaps).