Just came across this on slashdot. (http://developers.slashdot.org/developers/04/04/30/1650234.shtml?tid=141&tid=188)
Dirac is a general-purpose video codec aimed at resolutions from QCIF (180x144) to HDTV (1920x1080) progressive or interlaced. It uses wavelets, motion compensation and arithmetic coding and aims to be competitive with other state of the art codecs.
This post (http://developers.slashdot.org/comments.pl?sid=105976&cid=9021589) has some interesting information by someone who claims to be one of its developers.
http://sourceforge.net/projects/dirac (http://sourceforge.net/projects/dirac) / http://www.bbc.co.uk/rd/projects/dirac/overview.shtml (http://www.bbc.co.uk/rd/projects/dirac/overview.shtml)
Anyone got more information or actually tried it?
Wicked! BBC R&D up to some good quality work!
Now how long before ChrisHJW jumps on it?
"Matroska! Matroska! Matroska! *stamp* *stamp*"
Ruairi
http://www.hydrogenaudio.org/forums/index....ndpost&p=206921 (http://www.hydrogenaudio.org/forums/index.php?act=ST&f=49&t=21155&hl=&view=findpost&p=206921)
The visual studio workspaces are to compile libraries, not executables.
So, no Win32 executable ATM...
I guess these source are only directshow filters, an encoder and a decoder.
I guess these source are only directshow filters, an encoder and a decoder.
Nope, there are no calls to the DirectX SDK anywhere
It is certainly too early for them to worry about VCM and DirectShow filters, I was hoping for a command line app that would take uncompressed AVI or RAW YUV as input.
Nope, there are no calls to the DirectX SDK anywhere
True but there are main functions in encoder and decoder directories. So here we've got a ready to compile compressor/decompressor
just a quick related question, does "Dirac" has anything to do with the "Dirac Notation" from Quantum Mechanics? I seriously hate that guy for inventing that stuff ... dunno if I'll be able to use something with his name in it.
just a quick related question, does "Dirac" has anything to do with the "Dirac Notation" from Quantum Mechanics?
Yes, I would reckon it's a sort of homage to P. A. M. Dirac. He was british, so...
to add this too;
i hope this one's the first useable wavelet (althtough i doubt it'll happen...)
generally previous wavelet image and video compressors produced blur-fest stuff that nobody really wants...
No, it's not the first usable wavelet codec.
Rududu has been around for quite a while.
http://rududu.ifrance.com/rududu/ (http://rududu.ifrance.com/rududu/)
One of the BBC Research guys visits our workplace every week, i'll have to ask him about this.
It's a great idea and just what I want from the BBC - a decent use of my license fee. I can't make head or tail of some of the code though, i'll have to read up on wavelets.
It's quite sneaky of them really - with all the world moving to mpeg4, this is something for the future - the decode is far too slow for current computers - but when it does become viable on consumer devices, a *decent* open-source codec might cause a bit of a stir.
Wicked! BBC R&D up to some good quality work!
Now how long before ChrisHJW jumps on it?
"Matroska! Matroska! Matroska! *stamp* *stamp*"
Ruairi
It outputs raw frames ? Or does it have something like a native container we could parse in mkvmerge ?
p0l1m0rph1c i know what i said;i said useable....
rududu is not useable if you compare it to mpeg's....
it's blurred..it cannot achieve mpeg sharpness no matter what you do!
and i stay behind these words....
Yes, Rududu lacks the detail stored with MPEG based codecs, but it also lacks the incredibly irritating macroblock edges. Pick your poison. I wouldn't call it unusable, but I would say it has quite a way to go to mature. Definately promising for a one man side job.
Regardless, it would be nice to see this new codec turn into something great.
just a quick related question, does "Dirac" has anything to do with the "Dirac Notation" from Quantum Mechanics? I seriously hate that guy for inventing that stuff ... dunno if I'll be able to use something with his name in it.
perhaps they used the dirac delta function for solving the problem(s) so they named it after dirac. sure if you dont like dirac notation then use the mathematicians notation for inner products.
question: what sort of MC does it use?
rududu is wavelet intra coded, but block based motion compensation (so if you choose your settings very badly, you'll see blocks in rududu as well, though usually these are covered with intra wavelet stuff.)
i've heard that the collective "they" couldn't get 3d wavelets to work all that well for MC, so i'd be curious as to what BBC research have come up with.
(maybe i should read the article )
contrary to this.i say that main problem of wavelet is spatial performance (not ME/MC)
ie because of lousy intra it looks bluury,not because of block based inter....
i dunno what are we discussing here.....this codec doesn't exist as such....yet...
so it doesn't use anything,and it looks like nothing....
i dunno what are we discussing here.....this codec doesn't exist as such....yet...
Nic got it to encode in a Doom9 thread here. (http://forum.doom9.org/showthread.php?s=&threadid=75387)
ok,sorry...i've read nic's stuff now...
i correct myself;it exists but it's not quite working yet....(hehe)
when we have a full working order codec we can do some tests and perhaps help developers to make it better....
(although,if it stays "slower than h264",then i can't try it anyway...)
sorry for my scepticism about wavelets,but as i said,i hope this one will be exception...