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 > MPEG-4 AVC / H.264

Reply
 
Thread Tools Search this Thread Display Modes
Old 2nd September 2008, 14:43   #1  |  Link
Inventive Software
Turkey Machine
 
Join Date: Jan 2005
Location: Lowestoft, UK (but visit lots of places with bribes [beer])
Posts: 1,953
x264 - the state of play at the moment

I've been following development of x264 for the past few months without using it on a regular basis. I now have need to use it for transcoding VHS>DVD backup to H.264. There is one slight problem. The "Current patches...." thread has really confused things for me, and all the patches flying around at the moment are completely screwing with my head with regard to what's good, what doesn't work, what's completely broken, and what goes against all advice but still comes out on top.

As I understand it, Dark Shikari has a few patches in limbo (not completely finished, but he'd like them in the trunk) at the moment:
  • VAQ2
  • PsyRDO
  • PsyTrellis
  • New B-Frame decision

Correct me if I'm wrong, but this is how I follow things.

VAQ2 I was following based on the amazing picture quality of VAQ1, which eventually made it into the GIT trunk. PsyRDO has helped reached a level of confusion with people applying various patches to vanilla x264 builds that appear to clash with their intended operation, and from memory, VAQ2 is deprecated based on PsyRDO. PsyTrellis is a new-fangled motion helping bring more quality to various settings, and completely screwing the metrics over. The new B-Frame decision mode appears to have big speed issues, but is much better with regard to *most* content.

I'll say it now: those who use metrics to try and prove something with PsyRDO and PsyTrellis are wasting their time. "Psy" is short for "Psychovisual", very loosely meaning screw with the head via the eyes. People would do well to quote and say "Amen" to get the bloody message across!

I realise there are probably a few major bugfixes to make it into the GIT trunk in the short term, but what I'm asking is: what's a good build that I can archive stuff with? Ordinarily I'd use the x264.nl Sharktooth links to in "his" post, but the PsyRDO and PsyTrellis patches have me interested!
__________________
On Discworld it is clearly recognized that million-to-one chances happen 9 times out of 10. If the hero did not overcome huge odds, what would be the point? Terry Pratchett - The Science Of Discworld
Inventive Software is offline   Reply With Quote
Old 2nd September 2008, 15:05   #2  |  Link
burfadel
Registered User
 
Join Date: Aug 2006
Posts: 2,229
Isn't the PsyRDO and PsyTrellis part of the same patch?
Anyways, I guess the patches to choose are the ones that give the best subjective quality/file size for your intended purpose. The new b-frame decision patch and the psyrdo/trellis one are both good, just stick with 5 b-frames using the command line option --b-adapt 2

Trellis has to be set to either 1 or 2 for psytrellis to work (which kind of makes sense), although 2 would be a much better option that 1, especially considering a bugfix made to trellis 2 in a very recent build (the bugfix slows down encoding but gives better quality than before). Psyrdo requires a subme or 6 or greater, and works best with --b-rdo (makes sense also).

The easiest thing to do is try the rev 956 mod build as on the x264.nl homepage, all the useful x264 builds are very similar, but its great to have the choice! Ranguvar's, bobor's, skystrife's, x264.tk (and no doubt I'm missing a few) all make good builds, its kind of a matter of choosing the latest one with the patches that you want in them! sometimes they may omit certain patches for certain builds.

I wouldn't worry about using matrices files, x264 does a very good job, and they become redundant really when considering AQ, psytrellis, psyrdo etc etc. The b-frame decision patch is probably the one that needs the most work on it, in terms of performance and some anomalies, the others seem pretty good. They may just require a bit of final refining and testing before being committed. The target is to not introduce new bugs in to the official git, not to play catchup later!

Last edited by burfadel; 2nd September 2008 at 15:09.
burfadel is offline   Reply With Quote
Old 2nd September 2008, 16:54   #3  |  Link
Ranguvar
Registered User
 
Ranguvar's Avatar
 
Join Date: Feb 2007
Location: ::1
Posts: 1,236
He did not mean "matrices", as in custom quantization matrices, he meant "metrics" (SSIM and PSNR, among others) which are ways of roughly assigning a numerical value to encode quality. Useful for doing quick testing of patches and stuff, but useless for most "real" stuff.
Ranguvar is offline   Reply With Quote
Old 2nd September 2008, 17:02   #4  |  Link
Sharktooth
Mr. Sandman
 
Sharktooth's Avatar
 
Join Date: Sep 2003
Location: Haddonfield, IL
Posts: 11,768
most ppl confuse how metrics work and what metrics are. metrics measure differencies from the original image... that's completely different from quality. assuming metrics measure quality, is a big mistake.
Sharktooth is offline   Reply With Quote
Old 2nd September 2008, 17:06   #5  |  Link
wyti
Insane Encoder
 
wyti's Avatar
 
Join Date: Feb 2008
Location: Lausanne (Switzerland)
Posts: 142
No it's true if you disable all kind of psy optim.
Without psy optim the goal of a codec is to be as identical as possible from the source, and metric show us that.
__________________
Only two things are infinite, the universe and human stupidity, and I’m not sure about the former.
wyti is offline   Reply With Quote
Old 2nd September 2008, 17:10   #6  |  Link
Ranguvar
Registered User
 
Ranguvar's Avatar
 
Join Date: Feb 2007
Location: ::1
Posts: 1,236
Quote:
Originally Posted by Sharktooth View Post
most ppl confuse how metrics work and what metrics are. metrics measure differencies from the original image... that's completely different from quality. assuming metrics measure quality, is a big mistake.
True, and I know that. I was just saying what the general usage is, though I probably should've also mentioned what you said.

@wyti: In that case, metrics are more helpful, but still cannot replace a good visual test. Look at AQ, for example.
Ranguvar is offline   Reply With Quote
Reply

Tags
psyrdo, psytrellis, x264 builds, x264 patches

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 11:19.


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