Pages: [1]   Go Down

Author Topic: i1profiler and patchtool scrambled targets  (Read 2274 times)

smilem

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 297
    • Color Management Services
i1profiler and patchtool scrambled targets
« on: June 07, 2018, 01:55:07 am »

Hello, I made target in i1profiler and exported as cgats, then scrambled it in patchtool save as cgats.
If I load the patchtool scrambled patches in i1profiler I see not scrambled target preview!

Same file loaded in measuretool makes scrambled target. So how i1profiler descrambles the target?

« Last Edit: June 08, 2018, 10:39:12 am by smilem »
Logged

smilem

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 297
    • Color Management Services
Re: i1profiler and patchtool scrabled targets
« Reply #1 on: June 07, 2018, 03:41:28 am »

Found how to solve this, it's simple you have to Re-map ->"renumber the ID's"
The the target loads scrambled.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: i1profiler and patchtool scrabled targets
« Reply #2 on: June 07, 2018, 01:37:20 pm »

One of the issues I1Profiler has is that it can't average chart measurements by taking the same patch set and scrambling them differently. It can average successive measurement results by dragging and dropping the multiple measurement files but they have to have the same RGB sequences.

One generally gets much more dE variation from charts that a scrambled differently than the same chart measured repeatedly.

There are workarounds using patchtool but the simplest way to achieve averaging differently scrambled charts is to concatenate the RGB files first. I1Profiler will then average all the measurements that have the same RGB values.
Logged

smilem

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 297
    • Color Management Services
Re: i1profiler and patchtool scrabled targets
« Reply #3 on: June 07, 2018, 04:57:03 pm »

Thank you for this suggestion, what do you mean by saying:
"concatenate the RGB files first. I1Profiler will then average all the measurements that have the same RGB values."

How do I do this? what does "concatenate the RGB files" mean? Did you mean Join the files into one big file?
What about basiccolor improve?

Also what to use to scramble differently same chart?
« Last Edit: June 07, 2018, 05:03:21 pm by smilem »
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: i1profiler and patchtool scrabled targets
« Reply #4 on: June 07, 2018, 05:30:57 pm »

Thank you for this suggestion, what do you mean by saying:
"concatenate the RGB files first. I1Profiler will then average all the measurements that have the same RGB values."

How do I do this? what does "concatenate the RGB files" mean? Did you mean Join the files into one big file?
What about basiccolor improve?

Also what to use to scramble differently same chart?
You can use patchtool to generate differently scrambled patches of the same chart. Do it twice with the same chart. Patchtool also lets you add the two patch sets together to make one larger one.
Logged

smilem

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 297
    • Color Management Services
Re: i1profiler and patchtool scrabled targets
« Reply #5 on: June 08, 2018, 05:27:55 am »

Do you have any data on improvement in gray linearity with this approach?
I do simply rotate targets 180deg on laser printers to fix drum isues. But never tried the different scrambling of the same target for inkjets, is it really worth the trouble vs. 2500 gray patch target from Marc Levine?
« Last Edit: June 08, 2018, 06:00:50 am by smilem »
Logged

smilem

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 297
    • Color Management Services
Re: i1profiler and patchtool scrambled targets
« Reply #6 on: June 09, 2018, 08:20:25 am »

Quote
Do it twice with the same chart.

Well this works with the patchtool scramble, the Pollock and Vasarely produce same set every time  :o

The Patchtool has this PatchTool SUBSTRATE CORRECTION TOOL.
http://www.babelcolor.com/patchtool_substrate_correction.htm

This tool suposedly uses the method SCCA (Substrate Corrected Colorimetric Aims).

Why not ask the Patchtool author to make feature to virtually scramble the targets? Similar to the VPR for CURVE4.
This would need you to generate a target and have the cgats and tiff file handy. Then open the target tiff file in photoshop to delete all the patches resulting in empty target. Then you have to measure the empty target to get page substrate "fingerprint". Then you would print the normal target and measure again to get normal color patches measurements.

The patchtool having spectral measurements from empty page, full normal page could build a hotmap of the empty patches having most variantion. A list of this hot mapped empty patches would then be used as base onto which patch color only spectrum (paper minus color) would be simulated.

So one can expect to instruct the application to push all the patches trough this list of hot mapped empty patches. This would produce optimized set of data for building profile without printing multiple targets with different scrambling and measuring errors and wasted paper and time.
« Last Edit: June 09, 2018, 08:31:30 am by smilem »
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: i1profiler and patchtool scrambled targets
« Reply #7 on: June 09, 2018, 01:09:33 pm »

Well this works with the patchtool scramble, the Pollock and Vasarely produce same set every time  :o

The Patchtool has this PatchTool SUBSTRATE CORRECTION TOOL.
http://www.babelcolor.com/patchtool_substrate_correction.htm

This tool suposedly uses the method SCCA (Substrate Corrected Colorimetric Aims).

Why not ask the Patchtool author to make feature to virtually scramble the targets? Similar to the VPR for CURVE4.

It's already trivial for me to generate scrambled patches. For instance the following generates CGATs files, and associated iSis tif files for two, concatenated, scrambled rgb patch sets.

rgb1 = sortrows([rgb rand(length(rgb),1)],4);
rgb2 = sortrows([rgb rand(length(rgb),1)],4);
MakeIsis([rgb1(:,1:3); rgb2(:,1:3)], '2X target', 10.5, 7)

This creates both a CGATs file to load into the iSis patch set and associated, formatted and labeled tif file for printing the targets with the patch width set for 10.5mm and a 7mm height on US letter size sheet(s).

Quote
This would need you to generate a target and have the cgats and tiff file handy. Then open the target tiff file in photoshop to delete all the patches resulting in empty target. Then you have to measure the empty target to get page substrate "fingerprint". Then you would print the normal target and measure again to get normal color patches measurements.

The patchtool having spectral measurements from empty page, full normal page could build a hotmap of the empty patches having most variantion. A list of this hot mapped empty patches would then be used as base onto which patch color only spectrum (paper minus color) would be simulated.

So one can expect to instruct the application to push all the patches trough this list of hot mapped empty patches. This would produce optimized set of data for building profile without printing multiple targets with different scrambling and measuring errors and wasted paper and time.
I have explored variations due to various settings and the locations of patches of the same color as well as the paper white background. The former is a larger effect and the latter is not significant relative to other error sources.
Logged

smilem

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 297
    • Color Management Services
Re: i1profiler and patchtool scrambled targets
« Reply #8 on: June 09, 2018, 02:10:36 pm »

Quote
It's already trivial for me to generate scrambled patches. For instance the following generates CGATs files, and associated iSis tif files for two, concatenated, scrambled rgb patch sets.

rgb1 = sortrows([rgb rand(length(rgb),1)],4);
rgb2 = sortrows([rgb rand(length(rgb),1)],4);
MakeIsis([rgb1(:,1:3); rgb2(:,1:3)], '2X target', 10.5, 7)

This creates both a CGATs file to load into the iSis patch set and associated, formatted and labeled tif file for printing the targets with the patch width set for 10.5mm and a 7mm height on US letter size sheet(s).

Thank you for sharing this, how to use this? matlab is 13GB on 2 DVD's with dozen of toolboxes and stuff.
I never used it before.


Quote
I have explored variations due to various settings and the locations of patches of the same color as well as the paper white background. The former is a larger effect and the latter is not significant relative to other error sources.

You mean printing same color patches on various locations on the page has bigger influence then paper white difference in different locations?

But this is very strange to understand by logic, because printer needs to have stability in printing to be worth profiling. If the printer prints same color differently in measurable quantities and the substrate (paper white) is not related to this difference then the printing process is unstable. No?

If printing would be stable then why my proposed solution would not work?? If you have paper white substrate spectrum and patch color spectrum separated, you can simulate same color printing in any location but without need to actually print and scan. Curve 4 VPR shows this is reliable and working fine whether they use the SCCA (Substrate Corrected Colorimetric Aims) or other method is unknown to me.
« Last Edit: June 09, 2018, 02:27:49 pm by smilem »
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: i1profiler and patchtool scrambled targets
« Reply #9 on: June 09, 2018, 03:22:41 pm »

Thank you for sharing this, how to use this? matlab is 13GB on 2 DVD's with dozen of toolboxes and stuff.
I never used it before.
Matlab comes with quite a lot of functionality. Everything but the MakeIsis function in the 3 lines of code is built in. The MakeIsis function I made to do things like modify individual patches which was necessary to determine, by measurement, effective aperture. Matlab is fairly expensive but they came out with a home edition a few years back for individuals using it for non commercial purposes at a far lower cost. I have both a commercial and home version. The latter is current. The former I use on the very occasion commercial usage.
Quote

You mean printing same color patches on various locations on the page has bigger influence then paper white difference in different locations?

Yep. Likely because the way paper feeds in a printer results in imperfect print head to paper gaping.
Quote

But this is very strange to understand by logic, because printer needs to have stability in printing to be worth profiling. If the printer prints same color differently in measurable quantities and the substrate (paper white) is not related to this difference then the printing process is unstable. No?

If printing would be stable then why my proposed solution would not work?? If you have paper white substrate spectrum and patch color spectrum separated, you can simulate same color printing in any location but without need to actually print and scan. Curve 4 VPR shows this is reliable and working fine whether they use the SCCA (Substrate Corrected Colorimetric Aims) or other method is unknown to me.

The paper white is not affected by head gap spacing. There is some variation when scanned due to the paper not being perfectly flat but these are small and almost non existent with matte paper because it lacks specular variation.
Logged

smilem

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 297
    • Color Management Services
Re: i1profiler and patchtool scrambled targets
« Reply #10 on: June 09, 2018, 03:55:16 pm »

Quote
Yep. Likely because the way paper feeds in a printer results in imperfect print head to paper gaping.

This is possible I agree.

Quote
The paper white is not affected by head gap spacing. There is some variation when scanned due to the paper not being perfectly flat but these are small and almost non existent with matte paper because it lacks specular variation.

How many different scrambled version is enough, 2 or 3 etc?
I suppose you scramble some specific way? or random as code above?
Did you compare turning target 180 degrees vs. scrambling same target 2 different ways?
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: i1profiler and patchtool scrambled targets
« Reply #11 on: June 10, 2018, 12:06:50 am »

This is possible I agree.

How many different scrambled version is enough, 2 or 3 etc?
I suppose you scramble some specific way? or random as code above?
Did you compare turning target 180 degrees vs. scrambling same target 2 different ways?

How many duplications of patches is needed depends on the purpose. With the 9800 I've added a lot of redundant near neutrals around L* 15 and L* 90 because that is an area of large, fairly abrupt shifts. Most in the b* but some in a*, for printer profile targets. This is specific to the 9800 though and I've seen no evidence this applies to other printers. That aside, more patches near the neutral axis helps. The neutral axis is critical because very small changes in a* or b* cause large perceptual changes. You can have a dE2000 that is larger than dE1976 near neutrals. This is the opposite of saturated colors where dE2000 is almost always less than dE1976. So extra neutral patches is likely widely beneficial.

One area where I use a lot of redundant, randomly located patches, is to check profile accuracy. One standard set is a combination of the Colorchecker colors and evenly spaced neutral patches. My standard test chart has 11 duplications which reduces the expected error due to print location and paper variables by a factor of about 3.5.  The resulting averages are a much better indication of profile accuracy. This approach was particularly useful for optimizing patch sets profiling the 9800.

Note that the work I was doing was for making specialized machine vision charts for use in a manufacturing QA process. Not really needed for making good, repeatable, prints for humans.
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: i1profiler and patchtool scrambled targets
« Reply #12 on: June 10, 2018, 10:06:21 am »

How many duplications of patches is needed depends on the purpose.
I use em when profiling presses (digital or otherwise) over the sheet to account for print/press variation when creating targets for profiles and as importantly, process control on such devices. VERY useful and telling!
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: i1profiler and patchtool scrambled targets
« Reply #13 on: June 10, 2018, 01:38:07 pm »

I use em when profiling presses (digital or otherwise) over the sheet to account for print/press variation when creating targets for profiles and as importantly, process control on such devices. VERY useful and telling!
Absolutely!

This is similar to how I use large numbers of duplicate patches. The standard deviation distribution amongst the same patches proved extremely useful for optimizing the vacuum settings for different papers. I was able to reduce the Lab variance amongst identical patches by 50% from the defaults. The standard deviations of dE relative to average color went from .4 to .25 for some of the worst colors.
Logged
Pages: [1]   Go Up