Re: MeasureTool Compare question
Re: MeasureTool Compare question
- Subject: Re: MeasureTool Compare question
- From: Roger Breton <email@hidden>
- Date: Sun, 20 Mar 2005 17:05:39 -0500
Title: Re: MeasureTool Compare question
> Why not simply convert the Lab patches to your destination profile, then
> convert back to Lab and use that as your target Lab, print and measure the new
> chart and compare to the target?
Well, I finally see light at the end of this tunnel. With your help and that of others, I now have a system for testing output profile’s performance -- simply. I start with an evenly spaced Lab sample, 210 values in total (both inGamut and Out-of-Gamut) that I build with InDesign. I convert this sample AbsCol to my Epson4000 profile and print it (I could use any profile instead). Then I measure the resulting colors with my ICcolor. In parallel, through Graeme Gill’s excelllent argyll set of utilities (I could have alternatively used Marti Maria’s LittleCMS excellent utilities too — thank’s Ignacio), I generate what those 210 Lab colors are supposed to look like on my printer, using AbsCol again. This time it is done from the command line, completely outside of the world of pixels -- beginning to love this stuff. Then I convert back to Lab the CMYK values calculated by argyll's icclu to yield the ‘predicted’ Lab values.
When I put everything together in Excel, I get the following:
deltaE (ab)
starting Actual measured Predicted (icclu) |----------------------------
Lab Lab (ICcolor) Lab | Actual Argyll PO5
|------------- |------------------------|------------------------|---------|---------|--------
A1 | 60 -4 -40 | 59.90 -3.55 -39.32 | 59.92 -3.84 -40.34 | 0.82 | 0.38 | 0.11
A2 | 60 -12 -40 | 59.41 -11.04 -39.30 | 59.93 -11.73 -40.16 | 1.33 | 0.32 | 0.05
A3 | 60 -20 -40 | 60.01 -18.88 -40.01 | 60.08 -19.98 -39.95 | 1.12 | 0.10 | 0.10
..
Up to 210 patches.
As you can see, the starting Lab value coming from Adobe InDesign is converted AbsCol to the printer profile (at print time only). Then, I measure the target in PO5. This gives the ‘actual’ Lab colors, what going through all this color management nonsense has actually achieved. Ideally, we would want the actual colors to mirror the starting Lab values, right? So that’s what we find in the first column under DeltaE, the ‘actual’. A low DeltaE is generally prefered to a high DeltaE, nothing new here. But what happens when we come accross a high DeltaE value? Is this because we’re up against an out-of-gamut color or is it because we’re up against a crummy profiling system? To know the answer, I started to use PO5’s built-in Color Calculator, which essentially looks-up value an output profile and reports DeltaEs (ProfileMakerPro similar facility is ColorPicker). This quickly led me to the reassuring conclusion that DeltaE values ranging between 2~3 to 45 (and beyond) were predicted accurately by PO5. But I wanted a way to automate this so that I could quickly test any profile. Lucky for me, there were Graeme Gill’s nice set of (free) color utilities that readily accept keyboard input or input from a file. After a little Unix/OSX pain to set them up in Terminal (and later in iTerm), I was able to use that system to calculated the ‘predicted (icclu) Lab’ values out of my profile conversions. This is in principle the same idea, I believe, that Bruce Fraser exposes in his book called “the round trip”. Although I’m not doing this strictly on in-gamut colors and I’m not using ColorLab nor MeasureTool. Both methods have their merits. What was critical to me, from my point of view, in order to evaluate an output profile’s colorimetric performance, was to have a method of 1) verifying the quality of the prediction of a profile and 2) whether those predictions actually meant anything.
On to some other projects now...
Roger Breton | Laval, Canada | email@hidden
http://pages.infinit.net/graxx
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden