[FFmpeg-devel] [PATCH] Implement histogram equalization filter.

[William Yu]

2010/11/25 Michael Niedermayer <michaelni at gmx.at>:
> On Thu, Nov 25, 2010 at 01:59:06AM +0100, Stefano Sabatini wrote:
>> On date Sunday 2010-11-21 18:57:46 +0100, Michael Niedermayer encoded:
>> > On Sat, Nov 20, 2010 at 04:44:01PM +0100, Stefano Sabatini wrote:
>> > > ---
>> > > ?configure ? ? ? ? ? ? ? ?| ? ?1 +
>> > > ?libavfilter/Makefile ? ? | ? ?1 +
>> > > ?libavfilter/allfilters.c | ? ?1 +
>> > > ?libavfilter/vf_histeq.c ?| ?217 ++++++++++++++++++++++++++++++++++++++++++++++
>> > > ?4 files changed, 220 insertions(+), 0 deletions(-)
>> > > ?create mode 100644 libavfilter/vf_histeq.c
>> >
>> > if you confirmed that this is binary identical to donalds filter then ok
>>
>> Uhm no, VirtualDub works in RGB colorspace this in YUV colorspace. I
>
> I was unaware of there being significnat changes between
> your code and donalds code.
>
> What we need is porting existing filters without introducing bugs.
> The best way to do this is a wraper around libmpcodecs.
I have a question: In my mind, the filters in libmpcodecs seems depend
on other mplayer's codes.
If we make a wraper around libmpcodecs, how to eliminate those dependencies?

> If filters are ported one by one they must NOT be changed.
> Proposed improvments must be carefully tested and be seperate of the porting.
> Without testing things will quickly become significantly worse and buggier
> than a libmpcodecs wraper ever could have been.
> Also it likely is better to keep the original filter and a changed filter
> seperate so as it preserve the original. That way the changes need less strict
> testing and review as little is lost if your personal eq filter messes up
> chroma.
>
> Summary
> if we have 1 eq filter it must be carefully reviewed and tested.
> if we have a donald_eq filter that is supposed to be exactly donalds its much easier
> also if have a stefano_eq and a baptiste_eq theres no real need to test them,
> for each ot just would be the maintainers decission on what to put in there
> be that good or bad. But with a single eq filter that has to work as the user
> has no alternative -> thus we cannot change it without extensive tests
>
> i hope this makes sense as iam a bit tired already and should go sleeping;)
>
> [...]
> --
> Michael ? ? GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
>
> When you are offended at any man's fault, turn to yourself and study your
> own failings. Then you will forget your anger. -- Epictetus
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.10 (GNU/Linux)
>
> iEYEARECAAYFAkzt5mAACgkQYR7HhwQLD6sCYwCfb7qB5/53dgnQ8LbDesWMTHoW
> /pkAn2hUKq678Izxyo0xl39qetOEw752
> =Jt4X
> -----END PGP SIGNATURE-----
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at mplayerhq.hu
> https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel
>