[argyllcms] Re: Profile input white not mapping to output white

  • From: Ben Goren <ben@xxxxxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Fri, 23 Nov 2012 13:41:56 -0700

On 2012-11-23, at 11:36 AM, Iliah Borg wrote:

> On Nov 23, 2012, at 11:50 AM, Ben Goren wrote:
> 
>> I just looked a bit closer, and I can only see a half-dozen pixels clipped 
>> in the second green channel in patch N18, and they're all outside of the 
>> BOX_SHRINK margins.
> 
> Flare still hits around those, and optical cross-talk too. I would definitely 
> try to expose not that hot.

All of the clipped pixels are coming from a bit of the Teflon tape that's 
sticking up a bit at the one edge -- and, again, that edge is outside of the 
area that Argyll is sampling, and everything within the sample area is well 
below clipping. I suppose there could be a bit of flare and bloom going on, but 
I still don't think that's the problem.

> Look at white point diagnostics from the colprof, is the white rendered 
> neutral?

Yes, very much so. You can see it visually in my post that just crossed yours. 
Here's a trimmed example of the output from colprof:

> $ colprof -v 706C8559
> No of test patches = 660
> Estimating white point
> Picked white patch 408 with XYZ = 0.946830 0.983390 0.810050, Lab = 99.354158 
> -0.236898 0.093899
> Picked black patch 253 with XYZ = 0.000088 0.000085 0.000089, Lab = 0.077060 
> 0.022064 -0.034357
> Approximate White point XYZ = 0.946830 0.983390 0.810050, Lab = 99.354158 
> -0.236898 0.093899
> Creating optimised per channel curves
> Initial White Point XYZ 0.946830 0.983390 0.810050, Lab 99.354158 -0.236898 
> 0.093899
> [. . .]
> Doing White point fine tune:
> Before fine tune, rel WP = XYZ 0.909789 0.941809 0.768345, Lab 97.704832 
> 0.305168 0.722086
> After fine tune, rel WP = XYZ 0.962933 0.998631 0.823526, Lab 99.947049 
> 0.009190 0.019801
> Creating fast inverse input lookups
> White point XYZ = 0.893401 0.926164 0.754514, Lab = 97.071710 0.071686 
> 0.809596
> Find black point
> Black point XYZ = 0.001869 -0.000000 0.002479, Lab = -0.000092 7.547255 
> -4.679681
> [. . .]
> Profile check complete, peak err = 11.683588, avg err = 2.806606


Continuing:

>>> and general clipping on the darkest patch.
>> 
>> Actually, that's a good sign.
> 
> Maybe not. The black clipping in the camera is sort of unpredictable, there 
> is a problem with uncertainty and accuracy at low levels, and forcing that 
> all to be black may be not the best possible approach. Extrapolation often 
> works better.

Even still...the black is getting properly rendered in the output. It's only 
the highlights brighter than roughly R=G=B=200 that are problematic. Everything 
under that, including the Teflon tape (which is roughly R=G=B=140) is just fine.

So...to recap, the profile works great for the entire gamut of the target and a 
fair amount beyond it. It's only when it gets significantly brighter (but not 
darker) than the gamut of the target that it goes off the rails.

> dcraw is not a converter, it is a raw file decoder with some limited 
> conversion functions in it.


Even still, I think I might try seeing if I can get it to run on my rig just so 
I can eliminate the linear RAW processing as one more possible point of failure.

Cheers,

b&

Other related posts: