[vlc-devel] [PATCH 1/2] Sepia improve by being done in YUV

Jean-Baptiste Kempf jb at videolan.org
Sun Apr 10 14:35:54 CEST 2011


On Sun, Apr 10, 2011 at 01:10:48PM +0200, Branko Kokanovic wrote :
> On 09.04.2011. 20:05, Jean-Baptiste Kempf wrote:
> >> On Thu, Apr 07, 2011 at 11:58:12PM +0200, gamajun at seznam.cz wrote :
> >>> Bring major speed up for most videos
> Sorry for late answer. This patch is actually not optimization, it's a
> completely different algorithm for calculating sepia. As far as I'm
> concerned, it's good enough to be applied as it is,

Yes, it is a bit different, but we've made a few iteration with Martin
to be sure to not be too far from the old one.
Still, if you see fixes, or changes to take in account, please do so.

> * Maybe default value for intensity should be bigger - people will
> probably experiment with filters and we want to make good first
> impression. With your algorithm, I got best result with 130-150, 100 was
> too small (ok, this is mostly personal opionion only!)

This can be changed. I personally prefer 120 ;)

> * Intensity values->RGB conversion you do in RVSepia() function should
> use yuv_to_rgb() function from filter_picture.h. It's really cleaner and
> easier to maintain then for every video filter to have it's own
> conversion functions)
That is a fair point. Martin?

> It is different algorithm, but much, much faster!

Not to mention the SSE4 version ;)

Best Regards,

Jean-Baptiste Kempf
http://www.jbkempf.com/ - +33 672 704 734
Sent from my Electronic Device

More information about the vlc-devel mailing list