[argyllcms] Re: spec2cie

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 07 Oct 2013 17:54:50 +1100

Joe Tschudi wrote:


> The test with illumread has been made using a Grafilite
> (http://www.grafilite.com/product_info.html) in combination with an EFI proof 
> paper
> that contains almost no optical brigthener. So I am not sure why the results 
> come out
> so blueish when using specplot. But anway, if I use a third-party profiling 
> engine to
> use the custom illuminant from spotread (not illumread) I get a CCT of 5863K 
> which I
> presume is ok since it is what Grafilite is proposing with their budget light 
> box. With
> illumread something's odd, I also get 8096K with the external application - 
> maybe I
> made a mistake. But again, since I just need to measure the custom illuminant 
> and no
> OBC is present in the paper, spotread should go well for my purpose.

Illumread is intended for dealing with FWA containting papers. The results 
should be
very similar to spotread for non-FWA paper, so it's a puzzle as to why you got 
different results.

> But even when using the spotread result to convert the test.ti3 using 
> spec2cie, I get
> LAB 91.12 / -4.29 / -4.91, which is far more blueish than what the source 
> color looks
> like under a CCT 5863K light source (and the original color definitely looks 
> beige, not
> blue)

What it looks like can be very misleading, depending on what your vision has
adapted to.

> I don't really understand this topic deeply enough, so maybe my basic 
> presumption is
> wrong. How can I convert correctly a spectral reflectance curve to a custom 
> illuminant
> in order to get right LAB values (under this custom illuminant) for further 
> processing
> (printing, etc.)? In the end I want to see the right color under the right 
> light
> presuming I already did the ICC profile using the custom illuminant as my 
> white
> reference*

That's what "colprof -i illum.sp" is for. It's not at all clear why you think
you have a problem. The normal "Lab" values used in ICC profiling and
in the .ti3 files are all referenced to D50, so naturally an illuminant
with a CT of 5863K is blue compared to D50.

> *PS: I've used an i1Pro RevE using spotread (I presume it's M0) for all the 
> reflective
> and emissive meaurements. I did a profile using the CustomTest_spotread.sp ~ 
> 5863K as
> my white point using a third-party profiler successfully. Now I have TEST.ti3 
> and
> CustomTest_spotread.sp as my source color and my viewing light and I need to 
> convert
> TEST.ti3 to LAB in order to print the color (abs.col.) and view it under the 
> custom
> illuminant. Note that when doing all this with D50 default it's almost 
> perfect apart
> from the fact that the viewing light is slightly more blueish, so I need the 
> custom
> illuminant to correct this behaviour slightly (that is why I am assuming that 
> from
> 5000K to 5863K there can only be a slight change of LAB and not such a 
> drastic one as I
> am having with spec2cie).

If you are using some other profiler, then all bets are off - I've no
idea what it does.

With a typical relative colorimetric rendering, the absolute white is 
irrelevant -
white gets mapped to white. It's only when you use absolute rendering that it
becomes relevant. What exactly are you doing ? What is the ArgyllCMS workflow
(ie. colprof arguments, CMM option etc.) ?

> one observation of specplot to compare between illumread and spotread, find 
> attached the screenshot that shows left=spotread and right=illumread.
> If you analyze the curves they look practically identical except the range 
> between 480 and 570 in illumread (marked lighter between the blue guide 
> lines). 
> The part before and after look almost the same, they are just shifted. I 
> can't 
> recall correctly how I did both illumread and spotread but could that explain 
> why illumread is 8000K+ and if so could that be an error?

Yes, looks like a serious bug. I'll fix it in the next release.

Graeme Gill.    

Other related posts: