Pages: [1]   Go Down

Author Topic: Large set of optimal profiling patches  (Read 890 times)

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Large set of optimal profiling patches
« on: December 07, 2022, 04:19:08 pm »

Over the past 5 years or so I've experimented with various patch set configurations to determine what sets worked best with my 3 printers, 9500 II, Pro1000, and Epson9800. I used an i1isis 2 XL with page US letter 8.5x11" pages.

I've found a few printer anomalies. The 9500II patch color shifts a bit depending on the patches printed to either side. For instance a dark color tends to become darker/more saturated if the adjacent patches were similar. The Pro1000 tends to print differently on the first page than subsequent pages if it hasn't printed in a while. The biggest differences are after a time delay of a week but there are significant differences even after a few hours. The 9800 was the most stable but showed gradual shifts as the cartridges aged and ink levels decreased over a year. Likely due to gradual evaporation and ink density increase through the breather tubes. Still, remarkably stable. Even though the 9800 was most stable, the profiles needed to be somewhat larger for similar accuracy due to the printer's more lumpy RGB response. Especially on the neutral axis.

However, by creating a tool that combined patch sets, randomized them, and added extra patches to each page to detect page to page printer changes I was able to get good, consistent results and came up with quite good patch sets.

For smaller size patch sets (<1500 or so), embedding an inner grid produced much better profiles. As an example, an inner grid 4x4x4 inside an outter (main) grid of (5x5x5) produces 189 total patches and a side benefit is that the spacing between neutrals (R=G=B) is halved.

However, the advantage this has decreases gradually and at over about 1500 total patches disappears and the conventional, single grid performs as well or even better for larger patch counts.

For each of these approaches I also found a large improvement in neutral and near neutral color accuracy by adding a series of near neutrals RGB values. The main reason is that DeltaE 2000 is much more sensitive to small colors changes along the neutrals so extra patches significantly improves prints that have large areas of low/no saturation color.

I made large set of the RGB patches, both with inner grid and only the standard grid but all with extra near neutrals. The ones with the inner grid are labeled optN.txt where N is the total patch count and linN.txt for patches that only use the standard, outer grid. Also, scrambled versions are also provided. They have the exact same RGB patches but are randomized to reduce regional color shifts such as occur on my 9500II. These have a file name ending in 'x'. I prefer the scrambled but the unscrambled ones are easier to visually identify gross print errors such as selecting the wrong paper type.

Attached is a zip file containing these.
Logged

arobinson7547

  • Full Member
  • ***
  • Offline Offline
  • Posts: 187
Re: Large set of optimal profiling patches
« Reply #1 on: December 07, 2022, 04:46:29 pm »

Your research sure highlights the need to recalibrate back to 'some' / 'any' kind of reference. That's one of the advantages of a CMYK rips. But even the calibration routines in slightly newer machine (I know that on the Canon side the 8400 range had the ability to recalibrate BACK to the condition the Printer was in, hopefully before you profiled.

Thanks for the insight.
Logged

GWGill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 608
  • Author of ArgyllCMS & ArgyllPRO ColorMeter
    • ArgyllCMS
Re: Large set of optimal profiling patches
« Reply #2 on: December 07, 2022, 07:34:11 pm »

For smaller size patch sets (<1500 or so), embedding an inner grid produced much better profiles. As an example, an inner grid 4x4x4 inside an outter (main) grid of (5x5x5) produces 189 total patches and a side benefit is that the spacing between neutrals (R=G=B) is halved.
Sounds like Body Centered Cubic.
Graphics Gems V page 65 has a nice description of its properties in relation to geometric quantization.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Large set of optimal profiling patches
« Reply #3 on: December 10, 2022, 12:14:19 am »

Sounds like Body Centered Cubic.
Graphics Gems V page 65 has a nice description of its properties in relation to geometric quantization.

Yep. I believe Argyll has a generator option that does that too IIRC.

These were tested with i1Profiler. They were made to be 8 bit even values because i1Profiler has a not only can't handle fractional values as it produces images for profile printing that are only 8 bit which introduces errors albeit small.

My goto choices from this list are opt953x.txt for single page i1isis compatible prints and lin2849x.txt for 3 page profiles with 6x6mm patch sizes. The opt953x.txt is significantly better than the i1isis-i1profiler single page 957 patch set that comes with i1profiler.

For i1Pro2/3 profiles, choose the optnnnx.txt rgb set that fits the most patches into the number of pages one prints.

I have a program on github that can be used to evaluate one or more patch sets against an independent random set of colors and the neutral axis.

https://github.com/doug3236/i1Patches

Quote
In a single printing pass, create and verify accuracy of one or more ICC profiles from one or more sets of RGB profiling patches.

i1Patches Is a Windows based tool principally used to aggregate and randomize one or more, printer profiling patch sets along with randomly generated patches for independent accuracy verification. It produces tif and CGATs files compatible with XRite's i1Profiler and i1iSis or i1Pro2 spectrophotometers.

It is particularly good at removing externalities that strongly affect comparing different profile patch sets. These have been typically evaluated by printing charts, measuring patch color, making profiles, printing sets of known colors with these profiles, and finally, measuring these prints and comparing mesasured colors to patch RGB value colors.
« Last Edit: December 10, 2022, 12:22:50 am by Doug Gray »
Logged
Pages: [1]   Go Up