Pages: [1]   Go Down

Author Topic: Spectro measurements to print  (Read 2323 times)

delfalex

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 56
Spectro measurements to print
« on: April 30, 2018, 01:13:58 pm »

Hi,

Apologies if this or a similar topic has been discussed/ directed before.

Could anyone suggest a route / process I could take for getting as (visually) accurate result as possible - ideally with the least amount of manual data entry:

I have made a set of measurements of surfaces using a few different spectrophotometers which I want to visually compare the results of in print. The measurement format are in CIE L*ab & spectral using Basiccolor Catch's ISO flavour of CGATS.
There are about 40x measurements spread over 3x CGATS (per instrument).
I was thinking outputting via GMG proof.

My problem lies in not knowing the best way to get the (L*ab) data into a printable chart.

Has anyone made a similar test before before? Does anyone have any practical advice for a methodology?

Thanks

Alex
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Spectro measurements to print
« Reply #1 on: April 30, 2018, 02:34:21 pm »

Hi,

Apologies if this or a similar topic has been discussed/ directed before.

Could anyone suggest a route / process I could take for getting as (visually) accurate result as possible - ideally with the least amount of manual data entry:

I have made a set of measurements of surfaces using a few different spectrophotometers which I want to visually compare the results of in print. The measurement format are in CIE L*ab & spectral using Basiccolor Catch's ISO flavour of CGATS.
There are about 40x measurements spread over 3x CGATS (per instrument).
I was thinking outputting via GMG proof.

My problem lies in not knowing the best way to get the (L*ab) data into a printable chart.

Has anyone made a similar test before before? Does anyone have any practical advice for a methodology?

Thanks

Alex

To get a print color that matches a Lab color is quite easy if you have Photoshop and a profiled printer. In Photoshop you just do a color fill and select Lab as the value to fill. Then, to print it, you select Absolute Colorimetric. in the printer dialog if you have a standard, color managed setup. While you need to have a profiled printer. For most printers, you can use the profiles that come with your printer. You do not need a profile monitor to do this.

If you don't have Photoshop, you will need a program that lets you print a color using Absolute Colorimetric. Lightroom does not. I'm not sure about Gimp.

What are the Lab values you need to print? One of the limits is the printer's gamut. Some specialized logos or Pantone colors are outside the gamut most printers have. I can tell you whether that's an issue if you list the Lab values you need to reproduce.
« Last Edit: April 30, 2018, 02:40:09 pm by Doug Gray »
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Spectro measurements to print
« Reply #2 on: April 30, 2018, 03:35:23 pm »

Has anyone made a similar test before before? Does anyone have any practical advice for a methodology?
Depending on your goal, my advise is don't go there. Colorimetry and the dE testing is about color perception. It is not about color appearance. The reason why viewing a print is more valid than measuring it is because measurement is about comparing solid colors. Color appearance is about evaluating images and color in context which measurement devices can't provide. Colorimetry is about color perception. It is not about color appearance. Colorimetry based on solid  colors in very specific ambient and surround conditions.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Spectro measurements to print
« Reply #3 on: April 30, 2018, 03:53:55 pm »

Depending on your goal, my advise is don't go there. Colorimetry and the dE testing is about color perception. It is not about color appearance. The reason why viewing a print is more valid than measuring it is because measurement is about comparing solid colors. Color appearance is about evaluating images and color in context which measurement devices can't provide. Colorimetry is about color perception. It is not about color appearance. Colorimetry based on solid  colors in very specific ambient and surround conditions.

While a L*a*b* color can and will have different appearances depending on the context/image and illumination it appears in, the OP is trying to create a "chart" presumably printed with color patches that match various surfaces.

That is exactly what Abs. Col. is designed and works quite well for. I've used it extensively to match up various room wall colors with new paint over the years.

Interesting that he should be taking readings with an assortment of spectrophotometers. There is typically significant inter-instrument variation on the order of several dEs. So getting a visual match of a patch with a surface color may require some iteration.
Logged

delfalex

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 56
Re: Spectro measurements to print
« Reply #4 on: May 01, 2018, 08:23:26 am »

Thanks Rodney & Doug,

Rodney, agreed re perception vs appearance - the purpose of this comparison is inter instrumental rather than color appearance: various methods in lighting the samples brings a whole different level of appreciation to them and so am tackling that separately for the moment.

Doug, yes I'm trying to create a chart that can be printed as for it'll be easier to read than monitor softproofing;

- using Photoshop; which works with L*ab whole values - would you say this is accurate enough to show subtle differences?

- As Manually dialing in the values of 200-300 scans would be painful  - Are there any scripts that can extract the Lab data from the CGATS (maybe after an initial conversion to CSV), and transformed into an .ASE (L*ab color) Swatch Library?

-In terms of outputting to print would it be best to convert the L*ab color AbsCol to the Printer/Ink/paper RGB profile through Photoshop or would it be best to AbsCol convert it to a workspace RGB profile (i.e ProPhoto etc) and then send it to a RIP (for it to sort out AbsCol)?

In terms of printer gamut (will most probably be using an Epson 4900)

Two of the sample surfaces used were an Xrite SG & Basiccolor MesCal13 (see attached), I also used some tricky varnished paint samples like Ultramarine (i.e  for gamut limits: Spectrolino#1 - M3, L* 12.538, a 40.549, b -66.262) & Cobalt Blue (M3, L* 19.536, a 23.107, b -64.104).

Thanks for your time on this.
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Spectro measurements to print
« Reply #5 on: May 01, 2018, 09:05:48 am »

There are going to be differences between differing Spectros even of the same models although they should be smaller depending on the units and company. I’ve can dig up data comparing several iSis of the same rev, testing we did when we found larger than expected deltas between rev C and rev E models. And there lies just one example of difference in raw measurements between spectro’s of which there is nothing one can do! In this one example, a large client with dozens of units using them for QC found out the hard way how they can differ even from one “model” to another, this time due to hardware changes.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

delfalex

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 56
Re: Spectro measurements to print
« Reply #6 on: May 01, 2018, 11:33:54 am »

Absolutely - on screen the colour differences are quite obvious when going between manufacturer (& lessened between models & then units) hence why the testing and printing.

At this stage I can't see the inter-unit differences being too much problematic for the level of accuracy required - we'll see - rounding values up to the whole Lab number is quite a forgiving threshold of variation to start with - once I've figured out how it can easily done.


Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Spectro measurements to print
« Reply #7 on: May 01, 2018, 11:37:17 am »

Thanks Rodney & Doug,

Rodney, agreed re perception vs appearance - the purpose of this comparison is inter instrumental rather than color appearance: various methods in lighting the samples brings a whole different level of appreciation to them and so am tackling that separately for the moment.

Doug, yes I'm trying to create a chart that can be printed as for it'll be easier to read than monitor softproofing;

- using Photoshop; which works with L*ab whole values - would you say this is accurate enough to show subtle differences?
Quote
- As Manually dialing in the values of 200-300 scans would be painful  - Are there any scripts that can extract the Lab data from the CGATS (maybe after an initial conversion to CSV), and transformed into an .ASE (L*ab color) Swatch Library?
While you can't enter fractional Lab values you can display them. There's a rather oblique way in Photoshop. You can enter precision RGB values in a 32 bit RGB colorspace. But to do that with Lab values you will need to convert them with something like the precision calculator at brucelinbloom's site. It's a PITA but doable.

As for scripts, check out this Matlab toolset.
https://www.mathworks.com/matlabcentral/fileexchange/13788-optprop-a-color-properties-toolbox
Quote


-In terms of outputting to print would it be best to convert the L*ab color AbsCol to the Printer/Ink/paper RGB profile through Photoshop or would it be best to AbsCol convert it to a workspace RGB profile (i.e ProPhoto etc) and then send it to a RIP (for it to sort out AbsCol)?

In terms of printer gamut (will most probably be using an Epson 4900)

Two of the sample surfaces used were an Xrite SG & Basiccolor MesCal13 (see attached), I also used some tricky varnished paint samples like Ultramarine (i.e  for gamut limits: Spectrolino#1 - M3, L* 12.538, a 40.549, b -66.262) & Cobalt Blue (M3, L* 19.536, a 23.107, b -64.104).

Many problems here. M3 will not correlate well with visible differences in highly saturated colors. It's a specific tech. for measuring spectra by significantly reducing surface specular reflections. The visual appearance of colors will have them. And printers have a great deal of difficulty reproducing the more saturated M3 measured colors with these specular reflections, which otherwise reduce saturation.

As an aside, Andrew's points about inter-instrument variation is very relevant and I have seen very little work done in this area. They get away with this mostly because there is also a great deal of individual human variation in color appearance where the smaller wavelengths dominate as in colors like cobalt blue. So little is gained there. Also, spectros that use so called white leds as an illuminant have rapid changes in the shorter wavelengths so this introduces errors with the relatively crude 10nm spectro spacing between tungsten and led based spectros.

Quote


Thanks for your time on this.
Logged

delfalex

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 56
Re: Spectro measurements to print
« Reply #8 on: May 01, 2018, 12:22:02 pm »

Hi Doug,

Sorry just trying to understand: when you said many problems here, were you alluding to the specific nature of M3 filtration (extends the gamut because or reduced specular - hence an extreme), or/and to the subjects of possible output conversion processes, and the gamut of the SG / MesCal13 Lab values with the SP4900?

As Rodney pointed out about inter instrument variations and you yourself concerning illuminants (filtrations, let alone geometry) make a big difference, I agree - hence why the testing and wanting to print out some charts - there is very little published testing on this (other than finding suitable replacements for existing instrumentation), so I'm starting by collecting and collating data, with the eventual findings directing me to the Spectro/s that do what's required.

Thanks for the link to the MatLab scripts and the L*ab conversion route to 32bit RGB - what RGB model would you suggest?

Thanks for your time on this.

Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Spectro measurements to print
« Reply #9 on: May 01, 2018, 05:03:01 pm »

Hi Doug,

Sorry just trying to understand: when you said many problems here, were you alluding to the specific nature of M3 filtration (extends the gamut because or reduced specular - hence an extreme), or/and to the subjects of possible output conversion processes, and the gamut of the SG / MesCal13 Lab values with the SP4900?
M3 is designed to reduce/eliminate specular reflections. It's mostly of use with prepress because wet ink reflects too much light. So they use M3 to get a better measure on how ink will look after it dries. Saving time in press runs is pretty important.

The other use is on things with very rough, but glossy like surfaces such as some canvas. Surface indentations of more than 20 degrees from the flat will have specular reflections that get picked up by regular spectros that are 45/0 geometry. This doesn't happen on normal semigloss, pearl, luster etc. but I've seen it on some canvas that uses PK ink. It's most effective at getting better deep shadow measurements but at the cost of less accurate measurements.

This problem with M3 is that it results in measurements that are more saturated than the light that is normally reflected from the canvas that needs it due to the absence of specular reflect that is a normal part of viewing the canvas.

Quote

As Rodney pointed out about inter instrument variations and you yourself concerning illuminants (filtrations, let alone geometry) make a big difference, I agree - hence why the testing and wanting to print out some charts - there is very little published testing on this (other than finding suitable replacements for existing instrumentation), so I'm starting by collecting and collating data, with the eventual findings directing me to the Spectro/s that do what's required.
It's still unclear exactly what you are trying to do. My closest guess is to print something that provides a visual impression of the different instrument readings of the same color. That is, as series of patches that illustrate how one instrument's color measurement would vary from another.

Most instruments are quite consistent measuring neutral and near neutral colors if their reference white patches haven't gotten too old. But they become increasingly divergent as saturation increases. Fortunately, the visual sensitivity to color differences also decreases as saturation increases though there is some non-linearity in hue changes depending on color angle.

My comments about variation between individuals is that the Lab values are made from the so called "standard observer" from data measured in 1931 using a small number of people. There are significant differences between individuals and different spectra with the same Lab values using the standard observer functions can result in people seeing different colors. It's an effect that can be greater than the differences between spectrophotometers. It's also a larger effect between monitors using different backlight tech. because the light is formed from combining spectrally peaked colors.
Quote

Thanks for the link to the MatLab scripts and the L*ab conversion route to 32bit RGB - what RGB model would you suggest?

Thanks for your time on this.
Logged

GWGill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 608
  • Author of ArgyllCMS & ArgyllPRO ColorMeter
    • ArgyllCMS
Re: Spectro measurements to print
« Reply #10 on: May 02, 2018, 03:50:33 am »

I have made a set of measurements of surfaces using a few different spectrophotometers which I want to visually compare the results of in print. The measurement format are in CIE L*ab & spectral using Basiccolor Catch's ISO flavour of CGATS.
There are about 40x measurements spread over 3x CGATS (per instrument).
I was thinking outputting via GMG proof.
The thing you haven't mentioned, but is critical for getting such a match, is what light source(s) are being used to view the original and the print. Are they the same ? Are they different ? How close spectrally are they to D50 ?

Note that L*a*b* values from typical graphic arts instruments assume a D50 illuminant, and that a print/proofing system only offers a colorimetric reproduction of color, not a spectral reproduction. (What's the difference ? - the former will only match under the illuminant it was computed for - default D50. The latter will match under any illuminant.)

Plus lots of other factors - are the target colors in gamut ? Does the paper have FWA/OBE ? etc.

If the illuminant is not spectrally D50, then the process for getting a good match involves measuring the illuminant, converting the target colors to colorimetric values for that illuminant, and creating a profile for the proofing system that uses that illuminant rather than D50 in computing colorimetric (i.e. L*a*b*) values. And using absolute intent.

[ But this does depend on what sort of fidelity you are after. If it just has to be in the ballpark, then plugging the L*a*b* values into Photoshop may be enough. ]



Logged

delfalex

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 56
Re: Spectro measurements to print
« Reply #11 on: May 03, 2018, 04:06:48 pm »

Sorry for the delay in responding, has been a rather busy couple of days


It's still unclear exactly what you are trying to do. My closest guess is to print something that provides a visual impression of the different instrument readings of the same color. That is, as series of patches that illustrate how one instrument's color measurement would vary from another.

Doug,
Yep - I have made a set of measurements off of a set of surfaces using a few different spectrophotometers measuring each one . Initially I want to visually compare the results of in print; the measurement formats are in CIE L*ab XYZ & spectral using Basiccolor Catch's ISO flavour of a CGATS text file. To get an initial idea of similarities & differences I was thinking to make a chart of the various measurements in Photoshop from the nearest whole Lab values, print them out and compare them to each other and to the original surfaces in a D50 viewing station.

Most instruments are quite consistent measuring neutral and near neutral colors if their reference white patches haven't gotten too old. But they become increasingly divergent as saturation increases. Fortunately, the visual sensitivity to color differences also decreases as saturation increases though there is some non-linearity in hue changes depending on color angle.

Thanks, will keep this in mind

My comments about variation between individuals is that the Lab values are made from the so called "standard observer" from data measured in 1931 using a small number of people. There are significant differences between individuals and different spectra with the same Lab values using the standard observer functions can result in people seeing different colors. It's an effect that can be greater than the differences between spectrophotometers. It's also a larger effect between monitors using different backlight tech. because the light is formed from combining spectrally peaked colors.

Thanks for the cue - I think this would be the next stage where I start to compare via spectral or XYZ rather than Lab - this I'd have to do in software - with amount of of samples though (6x per surface (excluding M condition variations), I wanted to weed out the ones that were obviously perceptually different first in print.

There's a rather oblique way in Photoshop. You can enter precision RGB values in a 32 bit RGB colorspace. But to do that with Lab values you will need to convert them with something like the precision calculator at brucelinbloom's site.

Looking at BL calculator; what Adaption would be the most suitable for going Lab > RGB32 (as well as XYZ > RGB32) : Bradford?

The thing you haven't mentioned, but is critical for getting such a match, is what light source(s) are being used to view the original and the print. Are they the same ? Are they different ? How close spectrally are they to D50 ?

Note that L*a*b* values from typical graphic arts instruments assume a D50 illuminant, and that a print/proofing system only offers a colorimetric reproduction of color, not a spectral reproduction. (What's the difference ? - the former will only match under the illuminant it was computed for - default D50. The latter will match under any illuminant.)

GWGill,
Good points; so the original surface and test print would be viewed under the same "GTI D50" illumination

Plus lots of other factors - are the target colors in gamut ? Does the paper have FWA/OBE ? etc.
Was thinking of using either Epson or GMG standard proofing paper to print on - that and the printer's gamut (via the Colourst RIP) would be up against the target colours (one of the sets target colours and Xrite SG target has some very saturated colours) - I'd have to take such reproduced colours with a pinch of salt

If the illuminant is not spectrally D50, then the process for getting a good match involves measuring the illuminant, converting the target colors to colorimetric values for that illuminant, and creating a profile for the proofing system that uses that illuminant rather than D50 in computing colorimetric (i.e. L*a*b*) values. And using absolute intent.
Understood - I have had to do this quite often before for artists who have made work under Flourescent lighting and can't believe that a certain colour is what they painted when they see the resulting documentation image on screen (or in print under full spectrum & D50 lighting)


[ But this does depend on what sort of fidelity you are after. If it just has to be in the ballpark, then plugging the L*a*b* values into Photoshop may be enough. ]

For the initial viewing L*ab will give me a basic idea - I can then narrow things down and start looking at the contenders one by one on a spectral level afterwards.

Thanks all for your time on this.

Alex
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Spectro measurements to print
« Reply #12 on: May 03, 2018, 11:48:38 pm »

Looking at BL calculator; what Adaption would be the most suitable for going Lab > RGB32 (as well as XYZ > RGB32) : Bradford?
Leave the calculator defaults at Bradford and ref white of D50. That's what the ICC and Photoshop uses when going from Lab to/from an RGB working space. The fact a monitor's white is set differently, perhaps to D65, is not relevant here.
Logged
Pages: [1]   Go Up