[MPlayer-dev-eng] [PATCH] Coring filter

Rich Felker dalias at aerifal.cx
Mon Jun 27 15:30:59 CEST 2005


On Mon, Jun 27, 2005 at 07:49:19AM +0200, Jan Paul Schmidt wrote:
> 
> Am 27.06.2005 um 01:12 schrieb Rich Felker:
> 
> >On Sun, Jun 26, 2005 at 11:28:12PM +0200, Jan Paul Schmidt wrote:
> >>Hi there,
> >>
> >>this is my type of a coring filter for MPlayer/MEncoder. See
> >>attachments. vf_coring.c is the new filter file and vf_coring.diff the
> >>patches to integrate it into the source tree.
> >>
> >>Coring sets everything below a treshold to black. This coring filter
> >>only uses luminance to make the decision.
> >>
> >>Feedback is really appreciated, as this is my first encounter with the
> >>MPlayer/Mencoder filters.
> >>
> >>jps
> >
> >This filter's not bad as far as I can see, but I question how good the
> >results will be. You'd be better off using eq2 to adjust brightness
> >and contrast so as to clip off the values below a certain level.
> >Otherwise you'll introduce artifacts right at the cutoff threshold,
> >and even if they're not visible on your monitor they'll show up for
> >people with different brightness/gamma settings, especially on lcd
> >panels.
> 
> Well, it is a cheap filter. Coring is always cheap but fast. You may 
> get similar result with eq2, but not that easy and fast. To that you 

Actually it's presently faster than your filter because it's
optimized. Easy too once you work out the formula for what to set
brightness and contrast to in order to get a certain cutoff.

> know for sure everything is zeroed below the treshold, even if you 
> might not see it. Coring is for butchering the low levels - no doubt - 
> and as always, the quality depends an the user. To reduce artifacts 
> there is the black parameter that cores to the treshold instead to 
> zero. But this is more like a last resort.

This sounds like a much safer way to do it.

> I guess you are probably questioning the nature of coring :o)

Yeah, that's a good way to put it. :)

Rich




More information about the MPlayer-dev-eng mailing list