[MPlayer-dev-eng] [PATCH] synch to new x264 revision

[Guillaume POIRIER]

Hi,
Le jeu 23/09/2004 à 15:38, Jeff Clagg a écrit :
> On Thu, Sep 23, 2004 at 02:32:42PM +0200, Guillaume POIRIER wrote:
> > Based on your observations, here's a patch for manpage's x264 section.
> > 
> > I understand that it's maybe too soon to have them included in an
> > "official" document, but it's still better than nothing.
> 
> First, I'm surprised by this. I figured this subject is for
> DOCS/tech/encoding-tips.txt.

Yeah, that's what I figured also, but I see several problems with that
way of documenting:
1) Most users aren't likely to read it since it's an unusual place to
find informations. Plus, so many users don't read the documentation at
all. That way, we can reply to a dub question "man mplayer".
2) A huge number of users use packaged versions of mplayer (deb, rpm,
you name it), where I don't think DOCS/tech/encoding-tips.txt is
available.
3) Apart from encoding-tips.txt, all the docs in DOCS/tech/ seem to be
devs-oriented. Let's be consistent.
4) If those informations are available in the manage, there's maybe a
greater chance that should there be mistakes in it, some users will tell
us.


> But I like the fact that the x264 options are
> (mostly) documented verbosely. Perhaps the main reason many people don't
> use lavc is because they're confused about what the options do. Actually,
> most people who DO use it are confused.

I agree with you. You'd be surprised how hard it is to find _any_
documentation about lavc appart from the manpage (That's the Destiny
that Diego set to me ;-). No wonder why people ask for XviD support,
included me. ;-)

> Anyway, as long as we're on
> the topic of the x264 documentation, a lot of the credit for it must go
> to Loren Merritt.
> He pointed out several wrong things that were put in
> there by Bernhard Rosenkraenzer and, later, by me. Plus, he provided a
> lot of detail and explanation.

I should again thank again Loren for helping me out on my task to better
document lavc. ;-)


> > Could you please review, and give feedbacks about the tips that I
> > previously added to x264 section (e.g. fullinter and Anime material)?
> 
> I really have no idea if it's true or not. I have only one cartoon DVD
> available, and it's not representative, since the picture is awfully
> noisy and shaky (the original series is from the 1970's). I have done
> nearly nothing in the way of systematic testing. I do recall one case
> in which fullinter was found to reduce PSNR, and it was on animation
> source. However, this was on an older version of x264 and I have reason
> to believe it may be different now.

Ok, I'm gonna do a test-run of x264 with Romain Dolbeau's lavcCenetic on
an anime as soon as I can.
I think that the bad results you had with fullinter and this old cartoon
is due to the noise you had. fullinter is supposed to kick ass on large
clean areas of the same color, which is something quite common on recent
anime.


> Most of the time I've spent doing anything resembling serious testing,
> it's been with a certain clip from the movie The Matrix. Specifically,
> a 216-second long clip that begins at the start of chapter 9. I did some
> tests with fullinter and never managed to pin down a difference either
> in PSNR or subjective quality.

That makes sense to me: not a whole lot of large areas of the same
color, so it just slows down encoding.


> If you want to see my favorite result so far, check the ftp incoming
> directory, matrix_c9_h264.avi (I'm a little uneasy giving out links to
> dvd rips on a publicly archived mailing list).

AFAIK, if you don't include sound, and that you keep it short it should
be okay.
Anyway, I don't have broadband, so I can't have a look at it right now.

Regards,
Guillaume

-- 
Logic:
Q: Provided that war is a crime... what is a war crime?
A: That's a soldier who refuses to kill another one.