PDA

View Full Version : Slightly OT, is there any wavelet delivery codecs?



Gunleik Groven
01-11-2007, 02:04 AM
Is there any wavelet based competitors to the web delivery codecs out?

Gunleik

Martin Jäger
01-11-2007, 03:45 AM
hej gunleik!

i'm not sure if i understand your question - but as far as i know
there is two codecs in quicktime based on wavelet:

- pixlet (only on mac-platform)
- jpg2000

pixlet runs realtime on my machine (playback ;) ) but not jpg2000
which also takes hell of a time to compress.

martin

Graeme Nattress
01-11-2007, 07:18 AM
The JPEG2000 implement in quicktime is based on a very old JPEG2000 library that has been updated many times since. I'd not base any ideas you have about JPEG2000 on the quicktime implementation. However, the quality of the image from it is pretty good.

I think there are a few companies out there with JPEG2000 based codecs for movie playback, and of course, Cineform has their wavelet based codec.

However, I'd not class any of these as delivery codecs. The only one that's "delivery" is DCI spec JPEG2000 which needs a horribly beefy server to run, and is very expensive.

I don't see any reason why wavelets shouldn't make for a great web codec though. If anyone's got the time, then it could be an interesting little R&D exercise....

Graeme

P Andersson
01-11-2007, 07:27 AM
very interested in this
if I understand it right wavelets allow for dynamic scaling
and would allow for some great incorporation in flash movies

Graeme Nattress
01-11-2007, 07:31 AM
Dynamic scaling? Don't know what you mean by that.

You can extract out half and quarter resolution and so on out of the wavelets, but that's just it. You can't make it any other size bigger or smaller without using traditional scaling techniques.

Graeme

P Andersson
01-11-2007, 07:52 AM
hi graeme

perhaps dynamic scaling is not the correct term

when you open a jpg2000 image in a browser for instance, it follows the size of the window with much less artifacting then if you change the size of an equally compressed regular jpg

posted to support your r&d suggestion and wanted to learn about how to compress a video stream that can change depending on the size of the users screen while maintaining a small movie file that scales without much artifacting, ideally this codec would function in a flash movie

thanks

Graeme Nattress
01-11-2007, 08:20 AM
To scale without artifacting is a function of the scaling algorithm. I don't see it as a specific codec property unless you're getting into encoding bitmap images as vectors.

Graeme

P Andersson
01-11-2007, 08:55 AM
The scaling algorithm? Interesting.

So then we are talking about the scaling being done by what, the application that is carrying the codec? If the same codec is shown in Quicktime or VLC or Flash or some other way, it would scale, decompress and view differently, is this what you mean?

I really like the way that you can use flash to show layers of images and videos, and user interaction together with random layers. I have done it with images for awhile and want to get into random layering of videos, all streaming from a website.

Never got into scaling the images as it produced too much artifacts, but then from reading this site I started playing with jpg2000 and found they behaved differently and produced smoother results, haven't implemented and tested this fully yet. But they do require a bit more cpu or no?

So there, that is a headache and a half, streaming video in a flash site, layered with varying opacity, and being able to use the same file at different scales, this would then have to be tempered against the average users computer being able to run it without getting stuck in rendering.

All of these things can't be done successfully now perhaps, at least not in a satisfying scale, but it is where my head is at today.

Roxco
01-11-2007, 10:14 AM
We can never predict whether all we want will be built in, but everytime they revise it, we get more cool tools to play with.

Hopefully Peter Hoddie still sleeps well at night as he brought us quite a few new tools for Mac based NLE's.

I don't know who is staying up late these days...

Rosco

GlennChan
01-11-2007, 08:29 PM
I don't see any reason why wavelets shouldn't make for a great web codec though. If anyone's got the time, then it could be an interesting little R&D exercise....
BBC is doing exactly that with Dirac (open source project). Wikipedia has info on it, as does Sourceforge (you can get their specs there).

Graeme Nattress
01-11-2007, 08:33 PM
Wow Glenn, you seem to be everywhere today. I'd forgotten about Dirac. Has that actually got anywhere yet?

Graeme

GlennChan
01-14-2007, 07:33 PM
Last I heard (from reading the forum), they finalize the spec. They are working on coding the codec, working on the motion estimation code (slow/good tradeoffs) / trying different algorithms for it, and working on performance. But if you really want to know, check their Sourceforge project because I might be wrong.

Not in beta yet.

Gavin Greenwalt
01-15-2007, 04:14 PM
Isn't the DCI spec based on JP2k?

Also Bink is Wavelet based and they are very widely adopted.

Edit: Whoops just saw Graeme already hit point one.

But yeah.. Bink. They're very well established, give them a try. Best of all I think you can compile your movie into a runtime using the RAD tools. (No codec required for your client!)

Graeme Nattress
01-15-2007, 04:42 PM
DCI spec is based on JPEG2000, yes.

Graeme