[FFmpeg-devel] rectification filter

[Michael Niedermayer]

On Tue, Aug 12, 2014 at 12:52:45AM +0200, Daniel Oberhoff wrote:
> 
> Am 11.08.2014 um 17:39 schrieb Clément Bœsch <u at pkh.me>:
> 
> > On Mon, Aug 11, 2014 at 05:33:10PM +0200, Daniel Oberhoff wrote:
> >> 
> >> Am 11.08.2014 um 00:20 schrieb Michael Niedermayer <michaelni at gmx.at>:
> >> 
> >>> On Sun, Aug 10, 2014 at 08:36:37PM +0200, Daniel Oberhoff wrote:
> >>>> 
> >>>> 
> >>>> Von meinem iPhone gesendet
> >>>> 
> >>>>> Am 10.08.2014 um 17:50 schrieb Michael Niedermayer <michaelni at gmx.at>:
> >>>>> 
> >>>>>> On Sun, Aug 10, 2014 at 03:36:04PM +0200, Nicolas George wrote:
> >>>>>> Le tridi 23 thermidor, an CCXXII, Paul B Mahol a écrit :
> >>>>>>> It is not mandatory(but it would be nice) to add other methods to have
> >>>>>>> this filter included into libavfilter.
> >>>>>> 
> >>>>>> Is it really a good idea? We would end up with various interpolation /
> >>>>>> anti-aliasing algorithms implemented in each filter that needs it, none of
> >>>>>> them with the same set of architecture optimizations and each with its own
> >>>>>> set of bugs and misfeatures.
> >>>>> 
> >>>>>> Until someone comes up with a really satisfactory solution, I believe the
> >>>>>> simple solution of suggesting users to upscale before the filter and
> >>>>>> downscale after, using the optimized lswr scalers, is better.
> >>>>> 
> >>>>> thats not practical
> >>>>> a 1024x768 image would need to be upscaled to 262144x196608 to get
> >>>>> 8bit precission from a nearest neighbor resampler as basis
> >>>>> 
> >>>>> also see vf_perspective.c which supports bilinear and bicubic
> >>>>> interpolation, these surely could be shared and exist already.
> >>>>> 
> >>>>> of course we could push it with just nearest neighbor and work on that
> >>>>> later, but i dont think only nearest neighbor and leaving it at that
> >>>>> is a reasonable choice, its too poor quality wise
> >>>> 
> >>>> I would like to push now, and try to generate a refactoring patch possibly within the week. 
> >>> 
> >>> sure, feel free to push and send a pull request, or if you prefer
> >>> i can allso apply the patch if you post the latest verision
> >> 
> >> I am not sure about pull requests (would have to read up on how to do that properly) so rigjht now I would prefer it by patch:
> >> 
> >> 
> >> 
> >> Is that ok?
> >> 
> > 
> > Yes, please just re-submit the patch here.
> 
> Darn, forgot that attachements are scrubbed from the list (I had the patch attached). Will paste at the end.
> 
> > 
> > Sorry about the suggestion of adding FATE tests, I forgot the fact that
> > the filter was actually using floats.
> 
> No probs :)
> 
> From aa27195163da74c6e8f6a3b258f971f589d19aca Mon Sep 17 00:00:00 2001
> From: Daniel Oberhoff <daniel at danieloberhoff.de>
> Date: Mon, 28 Jul 2014 23:58:12 +0200
> Subject: [PATCH] avfilter: ported lenscorrection filter from frei0r

patch applied

if you would change it to use fixed point instead of floating point
calculatios then a regression test would be possible

Thanks

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

In a rich man's house there is no place to spit but his face.
-- Diogenes of Sinope
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 181 bytes
Desc: Digital signature
URL: <https://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20140812/4f8bf2d2/attachment.asc>