What does the future of video encoding look like?

[Zarxrax]

I've been watching video encoding advance since the days of RealVideo and Vivo. The progress since then has been nothing short of amazing!
Looking back at how far things have come has gotten me thinking about where things might be going. What will video encoding look like 20 years from now?

It seems like every new "generation" of encoding technology aims to try and get about 50% improvement over the previous one, at the expense of more computational complexity.
Are there already interesting ideas and theories out there which could give us further massive improvements, but they are just not feasible with today's technology? Or is it more a case of the technology itself being the driver for incremental improvements in this field?
Will we finally hit a wall if we continue to iterate on the same basic technology that we have been using since mpeg1?

Just some interesting questions that I thought might make for a fun discussion :)

[Sulik]

IMO, there are two areas that have big potential:
1. Current codecs are primarily designed for natural video, and don't do very well on rendered or mixed graphics -> this is an area that will almost certainly see significant improvements (particularly with the many cloud computing applications).
2. Improved post-processing, focusing on subjective quality. One can look at audio codecs which are way ahead of video in this area, mainly due to the much lower processing cost (KB/s for audio vs MB/s for video).

[cihub]

Concerning "where things might be going" in a long term, I'd recommend you to read more about MPEG and its standards (http://en.wikipedia.org/wiki/Moving_Picture_Experts_Group#Standards). It is obvious that there's some point after which compression of 2D images will hit a wall similar to that which modern CPU's have hit. At that time we'll have a "paradigm shift" (I've seen software to recreate 3D scene from a single 2D image, which as you might guess probably wasn't very good, but truly impressive), -- we'll store, at least part of "frames", in 3D scenes, i.e. in meshes with textures. Maybe, later on, when computers will be powerful enough, we'll be able to generate textures at runtime as well. Also, it may be that if meshes will be too high-res someone will invent a technique to store only diffs of standard meshes (after all most creatures in this universe have very similar structure). Of course, it is something that won't happen even in the next 10 years.

As to the more "realistic" next generation video encoding you should google "Daala".

[feisty2]

In the remote future, computers can "think, imagine..." like humans
We don't need the huge data stack video anymore
We just need some arbitrary hints to describe the content of the video
Computers will rebuild the video to actual image (imagine) and show us
:D

[StainlessS]

Anybody working on a HoloDeck ? I want one.

[LoRd_MuldeR]

Anybody working on a HoloDeck ? I want one.

Oculus Rift is probably as close as you can get to that, at this time ;)

[pandy]

Anybody working on a HoloDeck ? I want one.

Before holodeck we need to invent (thermo)nuclear reactor (fusion), then second matter-antimatter reactor and WARP drive then third "omnidirectional holo-diode"

https://books.google.com/books?id=po7406HGXQYC&pg=SA13-PA7-IA1&lpg=SA13-PA7-IA1&dq=omnidirectional+holo-diode#v=onepage&q=omnidirectional%20holo-diode&f=false

[Sparktank]

If we can get a HoloDeck, we can surely get J.A.R.V.I.S. as an OS.
Can you imagine, not only learning-AI, but also one with charisma and fully capable of helping you generate a new element in the periodic table?

Give JARVIS a couple years, and he'll have x26x and avisnyth/+ fully developed to the point where all we need to do is load a video, sit back and watch a fancy lazer show synchronizing to Pink Floyd.

Give us about 50 human years and we can have a JARVIS that won't laugh at us everytime we ask it what the answer is to Life, The Universe, and Everything.

That aside, I believe with Windows 10 creating a Windows-genocide on win7/8/8.1, we'll all be forced to update to current hardware "market" standards.
From there, a lot of developers will have no choice but to give in and try out the new rigs and create a new starting point from there.

Honestly, if it's 2025 and Windows XP turns into the Resistance fighting against a Terminator-esque invasion of Windows 15 Titanium, I'm going to stick my computer in a gas oven and give it a satirical suicide ritual.

While companies are gearing towards using HEVC/h265 for 4K blurays in the coming years, I'm still giong to use AVC/h264 since it's had the most time to develop.
I'll still expirement with HEVC/h265 retail sales.

I mean, the fact I'm still encoding to MPEG2/DVD for compatibility, can you just imagine using a 4K/HEVC BD as a source?
...although... when you downscale to such a resolution with such an old format, you probably won't be able to tell that much difference over a standard 1080p BD as a source.

Maybe I'll get myself a BD-burner this christmas.
or really give AVCHD a try.

You know what. Considering the future is right around the corner, there's a lot I can try out right now.
Even before those 4K/HEVC BD's come out on retail shelves.

I'm still quite literally in the past, that the future is quite hopeful to me for video encoding.
I'm not part of the frontlines for the new Frontier or anything, so once most people get their tests going and post results, I can gauge how comfortable I am with my comfort zone and edge my way towards maturely-developed new technologies.

For practical use, not just experimental.

If this were a chess board, count me out. I'm all over the place right now.
I'm more liable to say "king me" at this point. :p

The future looks fun.

[movmasty]

Mah! Just store one I frame every 500 frames with few hints, then reconstruct the video by cpu,

A 4k video at 50kBytes/s

Then 200 years of stasis waiting for telepatic video encoding.