Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Video Encoding > New and alternative video codecs

Reply
 
Thread Tools Search this Thread Display Modes
Old 13th November 2013, 10:05   #21  |  Link
Procrastinating
Registered User
 
Procrastinating's Avatar
 
Join Date: Aug 2013
Posts: 71
Zero, at this stage daala is almost entirely theoretical. To my knowledge there has not been a single draft released for it yet, where VP9 is finished and h265 is in the final drafting stages.

Don't expect anything for a couple years yet.
Procrastinating is offline   Reply With Quote
Old 13th November 2013, 10:37   #22  |  Link
JEEB
もこたんインしたお!
 
JEEB's Avatar
 
Join Date: Jan 2008
Location: Finland / Japan
Posts: 512
Quote:
Originally Posted by Procrastinating View Post
...and H.265 is in the final drafting stages...
Just FYI, H.265 has been a finished specification since April 2013. Range Extensions (>4:2:0 and >10bit support) as well as other additions to the first specification are what are still in draft, and there should be a final ballot on these early next year.

Funny enough, the ISO/IEC still hasn't finalized the same format on their side, but that seems like an issue with bureaucracy.

And yes, Daala is a research project, and you should not expect much from it other than very interesting ideas and test implementations. If we get a nice video format with quite a few new ideas along the road in a few years, that's just extra .
__________________
[I'm human, no debug]
JEEB is offline   Reply With Quote
Old 13th November 2013, 13:03   #23  |  Link
dapperdan
Registered User
 
Join Date: Aug 2009
Posts: 201
Quote:
Originally Posted by zerowalker View Post
How does it compare to VP9 and X265?
I donīt mean in real quality but in the state it is in currently.
In the video I posted earlier, someone asks "where are you today" in the Q&A (at about 39 mins) and they say it's "not as good as VP8 today" but they're currently more a collection of individual experiments which aren't yet integrated together into the main codebase.

The idea is to aim for beyond HEVC/VP9 by the time it's ready (possibly becoming VP10) in a couple of years. But they do have running code right now, it's not just theoretical:

http://git.xiph.org/?p=daala.git;a=shortlog

One of their key development strategies (which they also used for Opus audio) is to deploy in situations where the output is 'disposable' (e.g. Mumble and Skype for Opus) and use that to gain feedback on progress before the format becomes final. They're currently looking for projects where this strategy would work for video, I'm guessing Mozilla's WebRTC might be one of them.

Basically, don't expect them to follow the MPEG/ITU playbook for developing a codec, rather look at the IETF Internet Audio codec process (which eventually gave us Opus).

Last edited by dapperdan; 13th November 2013 at 13:15.
dapperdan is offline   Reply With Quote
Old 14th November 2013, 14:02   #24  |  Link
dapperdan
Registered User
 
Join Date: Aug 2009
Posts: 201
Some more (fairly raw) videos of discussion and presentations about Daala, here:

http://people.xiph.org/~tterribe/daala/coding_party2/

Last edited by dapperdan; 14th November 2013 at 14:09.
dapperdan is offline   Reply With Quote
Old 15th November 2013, 15:00   #25  |  Link
zerowalker
Registered User
 
Join Date: Jul 2011
Posts: 1,121
It seems i misunderstood a post, which i thought was saying that there has been practical tests, meaning an encoder using the concept has been used, which is how i asked the question.
But if itīs only theoretical, which i myself concluded last time i read about it, than thatīs a different story.

I find it impressive that there is now 3 competitive rivals on stage, even if this one is farthest away.
zerowalker is offline   Reply With Quote
Old 15th November 2013, 19:15   #26  |  Link
mandarinka
Registered User
 
mandarinka's Avatar
 
Join Date: Jan 2007
Posts: 729
They do have an experimental encoder they work with and test stuff on. During one of the coding parties they even bolted inter prediction on IIRC. However, it doesn't produce anything useful (well, nothing does when you compare to x264, but you know what I mean - it doesn't nearly compress well enough, yet).

You can compile it and test encoding with it. I don't know about anybody making a windows build for normal people though.

Edit: there is a guy in the irc channel #daala at freenode who is running tests of Sintel trailer with it, he can show you his results /I don't want to put links to his ftp here.../

Last edited by mandarinka; 15th November 2013 at 19:21.
mandarinka is offline   Reply With Quote
Old 15th November 2013, 19:20   #27  |  Link
zerowalker
Registered User
 
Join Date: Jul 2011
Posts: 1,121
Ah then i understood it correctly more or less.
But yeah as you say comparing head to head with x264 is impossible for anything at this stage, at least in normal situations. Low bitrate is another story, at least with vp9 from my tests.

But well doensīt seem like anything useful for me then, even if it would be fun to play around with. My compiling skills are non-existent;P
zerowalker is offline   Reply With Quote
Old 22nd November 2013, 13:49   #28  |  Link
mzso
Registered User
 
Join Date: Oct 2009
Posts: 930
If I understand correctly the "big thing" in Daala is some sort of overlapped block transformation. Isn't anything possible beyond that, which doesn't use blocks at all?
mzso is offline   Reply With Quote
Old 22nd November 2013, 15:02   #29  |  Link
LoRd_MuldeR
Software Developer
 
LoRd_MuldeR's Avatar
 
Join Date: Jun 2005
Location: Last House on Slunk Street
Posts: 13,248
Quote:
Originally Posted by mzso View Post
sn't anything possible beyond that, which doesn't use blocks at all?
Wavelet Compression

Though, AFAIK, Wavelets could be block-based as well and DCT doesn't necessarily have to use (small) blocks. It's just how they are used in real-world to exploit the individual strength of each type of transform.

See also:
http://x264dev.multimedia.cx/archives/317
__________________
Go to https://standforukraine.com/ to find legitimate Ukrainian Charities 🇺🇦✊

Last edited by LoRd_MuldeR; 22nd November 2013 at 15:06.
LoRd_MuldeR is offline   Reply With Quote
Old 22nd November 2013, 17:03   #30  |  Link
mzso
Registered User
 
Join Date: Oct 2009
Posts: 930
Quote:
Originally Posted by LoRd_MuldeR View Post
Wavelet Compression

Though, AFAIK, Wavelets could be block-based as well and DCT doesn't necessarily have to use (small) blocks. It's just how they are used in real-world to exploit the individual strength of each type of transform.

See also:
http://x264dev.multimedia.cx/archives/317
I'm assuming. It's not without a reason that DCT codecs don't have one frame-size block.

I read that actually about wavelets. It's not promising. Guess I should have asked "anything good".
mzso is offline   Reply With Quote
Old 22nd November 2013, 17:21   #31  |  Link
vivan
/人 ◕ ‿‿ ◕ 人\
 
Join Date: May 2011
Location: Russia
Posts: 643
There once was http://forum.doom9.org/showthread.php?t=116229 but it's kinda hard to understand how it actually works.
vivan is offline   Reply With Quote
Old 23rd November 2013, 01:42   #32  |  Link
LoRd_MuldeR
Software Developer
 
LoRd_MuldeR's Avatar
 
Join Date: Jun 2005
Location: Last House on Slunk Street
Posts: 13,248
Quote:
Originally Posted by mzso View Post
I read that actually about wavelets. It's not promising. Guess I should have asked "anything good".
Well, apparently nobody has come up with something practical that was clearly superior to the state-of-the-art block-based DCT compression schemes in the last two decades

Even HEVC/H.265, which is supposed to be the next big improvement in video compression, sticks with block-based DCT - though they made the macroblock sizes much more flexible (and now call them "coding tree units").

Maybe video compression has reached a point where things evolve gradually rather than using radically new concepts? SIF1 sounds interesting, but has anybody made a comprehensive comparison?
__________________
Go to https://standforukraine.com/ to find legitimate Ukrainian Charities 🇺🇦✊

Last edited by LoRd_MuldeR; 23rd November 2013 at 01:47.
LoRd_MuldeR is offline   Reply With Quote
Old 24th November 2013, 20:05   #33  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 4,750
Quote:
Originally Posted by LoRd_MuldeR View Post
Maybe video compression has reached a point where things evolve gradually rather than using radically new concepts? SIF1 sounds interesting, but has anybody made a comprehensive comparison?
Codecs are much more about the aggregation of refinements than any "one big idea." I think the HEVC folks said that no one new feature added even 1% of a compression efficiency improvement.

The challenge for radically different transforms is that they don't have to compete against basic DCT, but all the DCT refinements we've accumulated over 20+ years of codec development. So it's perfectly possible there is some fundamentally better alternative to block-based DCT-like encoding, but it'd take thousands of scientist/developer years to get it refined to a competitive place.

Daala's got some good ideas, though. Staying in the frequency domain for prediction and treating the output pixels as an out-of-loop rasterization for display is enormously promising. The broader approach of not thinking in output pixels so much could be very promising.

Variable-sized and shaped intra coded blocks is another thing I'd like to see more of.
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 1st December 2013, 12:47   #34  |  Link
foxyshadis
ангел смерти
 
foxyshadis's Avatar
 
Join Date: Nov 2004
Location: Lost
Posts: 9,558
AVC wouldn't have been possible in 1995, HEVC wouldn't have been possible in 2005; processing power gains contribute a huge amount to what's possible to put in a widely used standard. Likewise, some of the new advances are mostly for higher-def video, so that the encoding methods scale with the data. HEVC's original TMuC had lots of other cool stuff, like 64x64 DCT, rotated transforms (both primary and post-DCT), variable qpel algorithms, but the speed and complexity tradeoff just wasn't worth it. Someday they might be. (The rotated transforms especially would be extremely useful for a still-image format, since they're intra-only.)

Some incremental advances were to correct AVC's over-reliance on serial algorithms. Wavefront is the most obvious, but CABAC was heavily tweaked to be more parallel and longer-term (bigger dictionary in rar or 7-zip terms) instead of replacing it with one of the other suggested methods.

Aside from that, a huge amount of video research is heavily patented, and some useful inventions get dropped because the inventor demands too much or isn't willing to participate. Wavelets never gained much traction for this reason, they're a giant patent minefield. (To say nothing of curvelets, ridgelets, etc, which are all much better at modelling edges, but much more complex.) Because of the low uptake, there's also less research into how to make them nearly as fast as the modern DCT; it's a circular problem.

Daala's dropping of the "binary" part of BAC without using VLC is similar to TMuC's V2V entropy, and I've been eager to see what comes of it ever since it was described. That might be a free couple % gain right there, whereas most other advances require speed hits.
foxyshadis is offline   Reply With Quote
Old 1st December 2013, 18:09   #35  |  Link
MoSal
Registered User
 
Join Date: Jun 2013
Posts: 95
Quote:
Originally Posted by benwaggoner View Post
I think the HEVC folks said that no one new feature added even 1% of a compression efficiency improvement.
Does that apply to the 'larger block sizes' feature?!
MoSal is offline   Reply With Quote
Old 13th April 2014, 23:35   #36  |  Link
LigH
German doom9/Gleitz SuMo
 
LigH's Avatar
 
Join Date: Oct 2001
Location: Germany, rural Altmark
Posts: 6,752
I was not yet able to compile a Windows binary using MSYS/MinGW with my little experience. There are always some packages missing. One of the two MinGW environments I got available misses a package 'check'; I downloaded the sources and configured it, but make failed immediately...

Is there anyone else interested in trying to make an EXE?
__________________

New German Gleitz board
MediaFire: x264 | x265 | VPx | AOM | Xvid
LigH is offline   Reply With Quote
Old 14th April 2014, 06:35   #37  |  Link
qyot27
...?
 
qyot27's Avatar
 
Join Date: Nov 2005
Location: Florida
Posts: 1,419
I've been building snapshots from Daala's git repository every few months, whenever I think it's been a while since I last tested it (and yes, I do mean Windows binaries). You need to have libogg, SDL, and libpng, and a working autotools environment. Which, all things considered, makes it easier to do if you forgo MSys and just cross-compile everything. Build instructions for all that fun stuff are available in the tedious cross-compilation guide I wrote for mpv (Daala's not included in there because nothing other than its own test suite uses it). My general guess would be that the guide would *probably* work as-is under MSys, just remove any references to --build= --host= --cross-prefix= or whatnot in the instructions (although the hoops needed for SDL are just icky).

daala_r797-g377312e.7z

Daala build instructions:
Code:
Daala
=====

# libpng might not be necessary if --disable-dump-images is used
# SDL is required for the player.
# libogg is required to output ogg-encapsulated Daala streams.

git clone git://git.xiph.org/daala.git
cd daala
./autogen.sh
     PKG_CONFIG_PATH=/usr/i686-w64-mingw32/lib/pkgconfig \
     ./configure --prefix=$HOME/daala_build --disable-shared \
     --disable-unit-tests --disable-doc --disable-dump-images \
     --host=i686-w64-mingw32
make
make install
mkdir -p ~/daala_build/bin
cp examples/*.exe ~/daala_build/bin
cd ~/daala_build/bin
for n in *.exe ; do strip "$n" ; done
qyot27 is offline   Reply With Quote
Old 14th April 2014, 08:04   #38  |  Link
LigH
German doom9/Gleitz SuMo
 
LigH's Avatar
 
Join Date: Oct 2001
Location: Germany, rural Altmark
Posts: 6,752
Thank you for providing a build and advices. I'll see how far I'll get...
__________________

New German Gleitz board
MediaFire: x264 | x265 | VPx | AOM | Xvid
LigH is offline   Reply With Quote
Old 15th April 2014, 15:20   #39  |  Link
EncodedMango
Registered User
 
Join Date: Jun 2013
Posts: 65
Point a nab in the direction of the commandline arguments documentation?
EncodedMango is offline   Reply With Quote
Old 15th April 2014, 15:30   #40  |  Link
LigH
German doom9/Gleitz SuMo
 
LigH's Avatar
 
Join Date: Oct 2001
Location: Germany, rural Altmark
Posts: 6,752
There is very little.

encoder_example -h
Code:
Usage: encoder_example [options] video_file

Options:

  -o --output <filename.ogg>     file name for encoded output;
                                 If this option is not given, the
                                 compressed data is sent to stdout.

  -v --video-quality <n>         Daala quality selector from 0 to 511.
                                 511 yields the smallest files, but
                                 lowest video quality; 1 yields the
                                 highest quality, but large files;
                                 0 is lossless.

  -k --keyframe-rate <n>         Frequence of keyframes in output.

  -V --video-rate-target <n>     bitrate target for Daala video;
                                 use -v and not -V if at all possible,
                                 as -v gives higher quality for a given
                                 bitrate.

  -s --serial <n>                Specify a serial number for the stream.
 encoder_example accepts only uncompressed YUV4MPEG2 video.
The default mode is encoder_example -v 10, according to the tests I ran.

It is hardly optimized yet. The SDL player is not even able to play 640x272 in real time.
__________________

New German Gleitz board
MediaFire: x264 | x265 | VPx | AOM | Xvid
LigH is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 10:37.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.