Pages: [1]   Go Down

Author Topic: Messing around with tracked neutrals  (Read 631 times)

rasworth

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 357
Messing around with tracked neutrals
« on: May 13, 2019, 02:20:44 pm »

PatchTool is addictive when it come to experimenting with printer targets.

Started out with a simple Excel sheet to generate neutral L*a*b* values (see attached text file).  For want of a better scheme went with 50 L* values and 4 points per value - arbitrarily chose 2 as the basic a*b* increment, each set consists of [a*  b*] equal [cos(30)x2  sin(30)x2], [-cos(30)x2  sin(30)x2], [0  -2], [0  0].  The idea is to create three points at 120 degree intervals around the L* axis, and one in the center (see first graph).

The final step is to run the L*a*b* values thru an existing printer profile to create the tracked rgb values (see second graph).  So is this an adequate scheme to add 201 neutrals to a printer profile target?  More to the point, how does one tell without printing numerous targets and generating/testing profiles?  Is surrounding L* with three near-neutrals enough or is more better?  Is the neutral (0,0) point necessary?  Should the a*b* increment be larger?

Lots of questions, no apparent tools other than brute force testing to answer.  One thing I determined was the minimum cgats style wrapper necessary to bring a text file into both PatchTool and ColorThink Pro (per attached text file).

Richard Southworth
Logged

Alan Goldhammer

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3600
    • A Goldhammer Photography
Re: Messing around with tracked neutrals
« Reply #1 on: May 13, 2019, 04:57:10 pm »

I've been playing around with ArgyllCMS and the new control parameter Graeme introduced in the last update.  The targen utility that generates the targets can now add as many patches along the neutral axis as one deems necessary by just adding the number value in the command line, the '-n' command.  This and the other neutral weighting command '-N' allow one to refine a preconditioned profile.  You can use the Argyll command to generate a wire frame L*a*b* plot to see what the patch set looks like.  You can also import the '.ti1' file that targen generates in the Excel.  However the data values are not in L*a*b* but rather RGB data and XYZ.   I'm sure you could convert them to L*a*b* as needed if you wanted to.  I'm just content to see the patch distribution.

If you want to fool around with a data set, send me a PM as one cannot post Excel files.
Logged

rasworth

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 357
Re: Messing around with tracked neutrals
« Reply #2 on: May 13, 2019, 05:11:01 pm »

Alan,

How are the patches distributed?  All down the "center" or some nearby?

Richard Southworth

Added by edit - you can zip up any file and attach.
« Last Edit: May 13, 2019, 05:16:18 pm by rasworth »
Logged

Alan Goldhammer

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3600
    • A Goldhammer Photography
Re: Messing around with tracked neutrals
« Reply #3 on: May 13, 2019, 06:06:27 pm »

Cool, I didn't know you could zip things in Win10!!  The '-n' command puts however many patches you specify flat on the L* axis.  The '-N' command moves a certain percentage of the color patches closer to the neutral axis.  You need to look at the Argyll on-line guide to see how that works.  I specify '-n0.6' in the command.

File of the target spreadsheet and the patch lay out are in the zip file.
Logged

rasworth

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 357
Re: Messing around with tracked neutrals
« Reply #4 on: May 13, 2019, 08:22:06 pm »

Alan,

I downloaded the file, copied out the data, and viewed in ColorThink Pro.  It appears to me that there is a dense line of neutrals down the L* axis, but no surrounding values, at least not in a pattern I could discern.

Maybe Doug will check in with some wisdom as to the value of the surrounding near neutrals, and how they should be configured.

Richard Southworth
Logged

Alan Goldhammer

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3600
    • A Goldhammer Photography
Re: Messing around with tracked neutrals
« Reply #5 on: May 13, 2019, 09:42:12 pm »

The other file should show a visual of the patches
Logged

rasworth

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 357
Re: Messing around with tracked neutrals
« Reply #6 on: May 13, 2019, 10:08:21 pm »

Couldn't get the two file to link, extracted both into the same folder but the html file didn't do anything.

Richard Southworth
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1831
Re: Messing around with tracked neutrals
« Reply #7 on: May 13, 2019, 11:56:15 pm »

Funny. Evaluating tracking neutrals happens to be exactly what I was doing today. Here's what I did on the Pro1000 with glossy.

Create small, packed RGB array (M=4) which is 91 patches.
Create 3 different types of staggered neutral patches. The first was staggered neutrals along the device RGB axis at 15 units separation and produced 86 patches. The second was staggered neutrals along the Absolute neutral axis from a preliminary profile. These were made every 3 L* values reversed to RGB then a stagger value of 5 was added and an additional tracking neutral was added between each. This produced 132 patches. The third was the same as above, but along the paper white neutral. This is optimal for Rel. Col or Perc. printing which is what most do.

These patches were combined, duplicated, and scrambled. They fit on a single letter size sheet formatted for I1iSis 2.

The page was scanned and the RGB duplicates spectra averaged to reduce variance.

Then, the packed RGB set was extracted along with each of the three different near neutral sets into 3 CGATs files. These were dropped into the measurement tab in I1Profiler and the 3 profiles made for each. 1 staggered on device RGB and hence printer independent and the other 2 which were specific for printing either Rel. Col. or Abs. Col.

Then I took a set of L* from 3 to 100 in steps of 1. Each set was run through all 3 profiles converting to device RGB space for both Rel. Col. and Abs. Col. These RGB values were scrambled, printed, and scanned then de-scrambled.

I ran a statistical check on the results for the ideal Lab values v the scanned Lab values and calculated dE00.

Results:

The device RGB neutrals had:

Abs. Col. Ave dE00: .42   Worst 10%: .95
Rel. Col. Ave dE00 : .47   Worst 10%: .96


The Rel. Col Tracking RGB neutrals had:

Abs. Col. Ave dE00: .33   Worst 10%: .71
Rel. Col. Ave dE00 : .25   Worst 10%: .61


The Abs. Col Tracking RGB neutrals had:

Abs. Col. Ave dE00: .27   Worst 10%: .64
Rel. Col. Ave dE00 : .31   Worst 10%: .63


The first (total) ave numbers are quite consistent with run to run relative change from each other on the order of .02 or less. The 10% numbers have higher uncertainty because there are only 9 samples involved - probably a bit over +/-.05

So one can see that the tracking profiles that were made from Rel. Col. v Abs. Col. showed a small advantage when used to print the corresponding neutrals (either Rel. Col. (paper white) or Abs. Col. (Abs neutrals))

Thought I'd add the typical results I get from the 957 I1iSis default patch set. It consists of a 9x10x10 matrix where the red channel has 9 and the other channels have 10 grid points. The remaining 57 patches are a combination of neutrals and oddly staggered values.

The default set had an Ave neutrals of dE00: .54 and worst 10% of 1.40 with Abs. Col. so all three of these profiles are performing quite well with far smaller patch sets. Of course with only 91 color patches they are considerably worse. The color ave dE00 over the Colorchecker SG colored patches was .83 and worst 10% 1.8


These were made with M2 readings. Likely much bigger differences would occur with M0 or M1 since the paper is heavy on OBAs.

Edited to Add:

And indeed they do. With M1 profiles, selecting tracking profiles for either Rel. Col. or Abs. Col. offers significant advantages.

For M1 (Full uV D50) The Rel. Col Tracking RGB neutrals had:
The device RGB neutrals had:

Abs. Col. Ave dE00: .77   Worst 10%: .1.52
Rel. Col. Ave dE00 : .48   Worst 10%: 1.36


The Rel. Col Tracking RGB neutrals had:

Abs. Col. Ave dE00: .82   Worst 10%: 1.91
Rel. Col. Ave dE00 : .28   Worst 10%: 0.67


The Abs. Col Tracking RGB neutrals had:

Abs. Col. Ave dE00: .40   Worst 10%: 0.82
Rel. Col. Ave dE00 : .82   Worst 10%: 1.91



« Last Edit: May 14, 2019, 10:58:08 pm by Doug Gray »
Logged

Alan Goldhammer

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3600
    • A Goldhammer Photography
Re: Messing around with tracked neutrals
« Reply #8 on: May 14, 2019, 08:42:46 am »

Couldn't get the two file to link, extracted both into the same folder but the html file didn't do anything.

Richard Southworth
For some reason when you zip the html file it makes it unreadable.  I just tried to extract it from what I posted and you are right it doesn't do anything.  I also tried zipping and unzipping on my own PC and the same thing happens.  I don't know what's going on here.  I would try uploading it to my website for visualization but GoDaddy has some kind of glitch and my site is down right now.
Logged

rasworth

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 357
Re: Messing around with tracked neutrals
« Reply #9 on: May 14, 2019, 09:27:50 am »

Doug,

On one of your previous posts you referred to a stagger pattern but did not define.  I want to gain insight into which tracked stagger geometries are best, i.e. fewest neutral and near-neutral patches that result in "satisfactory" accuracy.  So what is the geometry you're using for your latest results?

Richard Southworth
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1831
Re: Messing around with tracked neutrals
« Reply #10 on: May 14, 2019, 10:29:52 am »

Doug,

On one of your previous posts you referred to a stagger pattern but did not define.  I want to gain insight into which tracked stagger geometries are best, i.e. fewest neutral and near-neutral patches that result in "satisfactory" accuracy.  So what is the geometry you're using for your latest results?

Richard Southworth

For device neutrals, step from 0 to 255 in 15 increments. For each step add additional triplets where each channel is increased by 15. For instance:

0,0,0, 15,0,0  0,15,0  0,0,15    15,15,15  30,15,15  15,30,15  15,15,30,  ...

Theory is similar to your 120 degree ab rotation. Increase in red channel is less cyan, green is less magenta, and so forth.

This works well with a small, but significant population of color patches. For that test I used 91 which is a 4x4x4 with an interspersed 3x3x3 set.

For the tracked sets, I started with L* every 3 units. Converted to RGB using either Rel. Col or Abs Col. Then added a smaller stagger value of 5 instead of 15 since the RGB set has already been mapped to very near the desired positions.


These are, however, the first tests I've done with simultaneous Rel Col and Abs Col on a single print run which eliminates subtle effects from the printer such as the occasional "cartridge agitation."

Added: Probably worth noting that every print is measured forwards and backwards, then the results are compared and averaged. This reduces spectro variations and also provides a mechanism to detect anomalous readings from dust contamination of low L* patches.
« Last Edit: May 14, 2019, 10:57:44 am by Doug Gray »
Logged

rasworth

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 357
Re: Messing around with tracked neutrals
« Reply #11 on: May 14, 2019, 10:56:16 am »

Doug,

Got it, thanks.

Richard Southworth
Logged

Alan Goldhammer

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3600
    • A Goldhammer Photography
Re: Messing around with tracked neutrals
« Reply #12 on: May 14, 2019, 03:09:29 pm »

I cannot post the html page of the patch set so you can see the distribution of patches.  Here is the Argyll targen command if you want to play around with it.  just change the profile name to something you have.  It generates a 1848 set of patches in a  'Patch_Set.ti1' file.  '-n51' puts 51 neutral patches right on the L* axis and the '-N0.6' command adjusts colored patches towards the neutral axis to improve colors in that region.  '-w' creates the web view of the L*a*b* patch set that can be rotated around in your browser.

targen -v -d2 -G -n51 -N0.6 -w -cPre_Cond_Profile.icm -f1848 Patch_Set
Logged

rasworth

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 357
Re: Messing around with tracked neutrals
« Reply #13 on: May 14, 2019, 04:46:28 pm »

Alan,
I copied it out of excel and viewed with ColorThink Pro, no problem.

Richard S.
Logged

Alan Goldhammer

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3600
    • A Goldhammer Photography
Re: Messing around with tracked neutrals
« Reply #14 on: May 14, 2019, 05:18:57 pm »

Alan,
I copied it out of excel and viewed with ColorThink Pro, no problem.

Richard S.
Great!  You should be able to see the patch distribution and the added neutrals
Logged

GWGill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 534
  • Author of ArgyllCMS & ArgyllPRO ColorMeter
    • ArgyllCMS
Re: Messing around with tracked neutrals
« Reply #15 on: May 14, 2019, 10:02:15 pm »

For some reason when you zip the html file it makes it unreadable.
If this is a x3d.html file, then you may have to bundle the x3dom.css and x3dom.js files with it.

See the documentation.
Logged

Alan Goldhammer

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3600
    • A Goldhammer Photography
Re: Messing around with tracked neutrals
« Reply #16 on: May 15, 2019, 08:42:29 am »

If this is a x3d.html file, then you may have to bundle the x3dom.css and x3dom.js files with it.

See the documentation.
Aha!!!  Yes, that's correct. 

EDIT:  I've packaged the two files that will allow those interested to see the L*a*b* profile patches for Moab Entrada.  these are the ones I used to create my final profile.
« Last Edit: May 15, 2019, 08:51:39 am by Alan Goldhammer »
Logged
Pages: [1]   Go Up