evolution of compression

[Hrvoje Simic]

I've recently seen Disney/Pixar's "Ratatouille" trailer in 1080p and for the 10Mb/s the picture was pretty good. H.264 really does wonders.

How low can this go (IPTV purpose speaking) in terms of codec development. What are the boundaries and how far are we?

I guess that's a question for you, Graeme.

[Graeme Nattress]

A totally simple picture of a blank screen can be represented by one little bit of data. If you add movement and shapes, this goes up, as if you add any kind of noise, or detail.

How near are we to the bounds - getting closer all the time. Issue being that if you watch it on a CRT it might look fine, but on a higher resolution, less rose tinted glasses display, you'll see more artifacts.

Graeme

[Hrvoje Simic]

lol
Well I know that much.....:o

I've watched it on Apple's lcd in actual size and no major artifacts there. Pretty good result for IP distribution to home users, though.

[Ken K]

How low can this go (IPTV purpose speaking) in terms of codec development. What are the boundaries and how far are we?

As long as processing power increases, we will keep seeing more codec development. Just as newer codecs like H.264 would have a hard time running on 10 year old hardware, codecs that come out in 10 years will probably have a hard time running on our current hardware.

But just think, we'll be viewing the 4k we create this year in 10 years and it will probably still be considered really high res. :D

Red: Forward Thinking

[Graeme Nattress]

Really this is a question in information theory, and also on the nature of "what is random".

Processing power helps, but the bounds of compression are related to the information content of the image sequence.

I'm downloading the trailer now, so I'll see!

Graeme

[Graeme Nattress]

Um.... I notice a lot of banding in gradients....

Graeme

[Graeme Nattress]

Just watching the Harry Potter one, and if you look at the opening with the smoke and the river Thames, there's a fair bit of artifacting there....

Graeme

[Ken K]

Really this is a question in information theory, and also on the nature of "what is random".

Processing power helps, but the bounds of compression are related to the information content of the image sequence.

Ahh yes, I see what you meant now. Yes, given the same information content, they'll hit that point. I was thinking of compression in general, assuming we'll see things like increased resolution, audio quality, etc.

I love pondering information theory. :D

[Graeme Nattress]

Increased resolution - yes, and bit depth too!

Now, where advances could come is with lossy compression, you get to "invent" plausible detail. I think there's still work to be done there, but we're into vast amounts of processing to do so. Again, you'll come up against a limit, but I don't think we've really started down this route too much yet.

Graeme

[Hrvoje Simic]

Is it me or is the compression much more visible on the film, rather than animations?

[Graeme Nattress]

That's noise and grain for you. It tends to hide posterisation, give false appearance of sharpness, and also cause artifacts as it's too complex to easily compress. We're then back to the nature of "random".

Graeme

[Hrvoje Simic]

Yep.
Got it. Thanks, Graeme.

[Hrvoje Simic]

That's noise and grain for you.
Graeme

I suppose you wanted to stress out the complexity of compressing noise and grain by that.

[Graeme Nattress]

There's two things, the complexity you mention, and also the fact that grain hides artifacts and also has beneficial effects on sharpness. It's a complex equation.

Graeme

[Steve Gibby]

Maybe a bit of the Chaos Theory of physics involved...

Somewhat like the title of the Moody Blues album "A Question of Balance"