Pages: [1] 2   Go Down

Author Topic: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals  (Read 7056 times)

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197

I've been working to reduce/eliminate subtle color and especially hue shifts that I sometimes see on prints. Especially on the 9800 which has some interesting traits. Unlike my Canon 9500 II, it is very, very good at printing colors that change little. Even after days of no use. But it has some adverse traits as well. The color it prints varies somewhat with the position of that color on a page. And it has some unpleasant, moderately abrupt changes in color when printing neutrals and near neutrals. This largely applies to a* and b* rather than L* and seems to be a shift in colored ink mixing that occurs with small changes in L*. OTOH, the 9500 prints much less consistent colors with shifts of up to 2 dE2k as the ink depletes but also from a day or two of non-use. It also has an interesting history effect. The color of a printed patch depends on the colors of patches printed on the same row. So perhaps there is some temperature shift occurring or other effects altering the colors. But it is a marked affect most noticeable in dark colors. A dark patch that is surrounded by other similar patches tends to be printed slightly more darkly. I do not see this with the 9800. However, the 9500 prints a much smoother full neutral gradient with less rapid variation on each of the 3 Lab axes, L*, a* and b*. So it doesn't benefit as much from a larger patch count or higher resolution profile LUTs than does the 9800.

Lest I complain too much, these are fairly subtle effects. Overall, prints made that are in gamut with both printers look identical outside of slight specular differences from bronzing and reflectivity and those don't occur at all with matte paper.

So, specifically to get the best neutral tone curves for B&W or color printing with a lot of near neutrals, I made a profile target that fully populated the neutrals and near neutrals while including an additional 1914 patches from the I1Profiler Optimize process. The fully populated neutrals and near neutrals were picked so that every A2B1, 3D LUT grid point in I1Profiler's max quality profiles had an associated RGB triplet and measured color. The result was a 4 page iSis target set with just under 3,828 patches

This produced excellent results. Both in more saturated colors and along the neutral axis.  Ave dE2k along the neutral axis was under .4 and under .5 for the regular dE76 with maxes under 1.

After seeing those good results I created an expanded near neutral set fully populating not only adjacent neutral grid points but all grid points within 3 locations along each of the three axes.

To this I added a set of RGB values covering all the rest of the RGB values such that every third 3D LUT grid point had an associated RGB triplet. While this doesn't provide as fine a set of colors for those off neutral and near neutral, they are aligned with the I1Profiler's technology when using max quality profile generation. And they should minimize interpolation errors when going from A2B1 which is used for soft proofing.

The result is a set of 2871 patches which creates 3 pages of standard iSis letter size targets. These can also be used for non-iSis charts by importing the CGATs file (Randomized I1P 12x cube etc.txt) and turning them into convenient charts in the desired physical format.

I have included a zip file with two sets, regular, and randomized. I mostly prefer to make profiles from the randomized charts. Especially with the 9500. Some prefer the regularity of standard charts because it can be easier to spot gross defects.

There are a few additional features in these charts. They are designed to be read upside down by an iSis and have the necessary reference bars on both the top and bottom. Rather than read the same chart in the same way twice, I read them in reverse order and backwards then combine the results with a utility that averages the results. I get small improvements doing this. If you wish to read them in reverse order and backwards load the second CGATs file to make the iSis targets.

For those with an iSis, use the ready made targets suffixed ".txf"

Randomized I1P 12x cube etc.txt  -> CGATs for Standard Isis Target
Randomized I1P 12x cube etcr.txt ->  CGATs for Reverse Order, Backwards Read Isis Target

One note. You can't automatically average the fwd/rev files when creating a profile like you can selecting multiple measurement files and dragging them in. They require the entries to be in the same order. They can be re-ordered if you are handy with Excel or conversant with PatchTool. I use a custom Matlab function to automate this.


« Last Edit: April 10, 2018, 07:38:37 pm by Doug Gray »
Logged

Jeremy Roussak

  • Administrator
  • Sr. Member
  • *****
  • Offline Offline
  • Posts: 8961
    • site
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #1 on: April 24, 2018, 01:42:44 pm »

I have deleted all the messages in this thread save for Doug's opening gambit.

Let's start again, shall we?

Jeremy
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #2 on: April 24, 2018, 01:56:06 pm »

I have deleted all the messages in this thread save for Doug's opening gambit.

Let's start again, shall we?

Jeremy
So the data point I made about original and optimization targets, isn't valid?
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Jeremy Roussak

  • Administrator
  • Sr. Member
  • *****
  • Offline Offline
  • Posts: 8961
    • site
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #3 on: April 24, 2018, 02:19:43 pm »

So the data point I made about original and optimization targets, isn't valid?

It might have been: I'm not qualified to say. But it was caught up in the spam. Feel free to repost.

Jeremy
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #4 on: April 24, 2018, 10:45:48 pm »

So the data point I made about original and optimization targets, isn't valid?

Andrew,

I recall you mentioned you used Bill A's set augmented with optimization patches from I1Profiler.

I used Bill A's set back around 2006 when I first got my 9800, but also a 2400 I had at the time. I used the older I1Pro with GMB's PM5 software. IIRC, Bill optimized it initially for the 9600, presumably with PM5 as well. PM5 was/is a good product and let you easily run batches but I1Profiler supports a finer resolution 3DLUT set. It's got some strange quirks but makes good profiles. Probably the oddest aspect of I1Profiler is the way it treats fractional RGB values. Sometimes it rounds, sometimes it truncates. So it's best to round everything to 8 bit RGB before making targets and profiles from custom patch sets.

Bill A. put a lot of work into his profile patch set but, since getting an iSis XL, I've gotten better results with custom profiles on the 9800 when combined with I1P. I find the 9800 is very (still after all these years) consistent and I1 Profiler generates really good profiles for it when loaded with extra near neutrals and I1P optimization patches.

I would expect your approach of using Bill A's patches and additional optimization patches would produce really good profiles for your printers. And things seem to have improved since those days of old.
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #5 on: April 24, 2018, 10:59:58 pm »

Bill's target, my own optimization target, correct.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #6 on: April 25, 2018, 11:54:28 am »

Doug, if you don't have this .pxf file, might be useful. Made by former X-rite employee Marc Levine and it contains 2500+ neutrals and near neutrals.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #7 on: April 25, 2018, 01:43:49 pm »

Doug, if you don't have this .pxf file, might be useful. Made by former X-rite employee Marc Levine and it contains 2500+ neutrals and near neutrals.

Interesting patch set. When loaded there are a few scattered strong colors which disappear when switching to the printable chart.  These are the result of byte wrapping since the RGB values go lower than 0, down to -3 and up to 260. They are clipped to 0 and 255 when printed.

Likely makes a really good B&W profile, especially for more modern printers than my 9800.  I don't think the close RGB values negate completely the 9800's magenta shift around L*90. Depends on whether it extrapolates well, not just interpolates.

I use a more limited set of near neutrals that focus on my printer's area of rapid change. Since they are specific to the printer they are a much smaller set. I have always combined them the larger set of color patches. Got it down to a 2 pg. set (1914 patches) for the 9800. Not quite as good as the 4 pg. set but very close.
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #8 on: April 25, 2018, 02:04:18 pm »

Interesting patch set. When loaded there are a few scattered strong colors which disappear when switching to the printable chart.
Yes, but if you toggle to the print view, they disappear. And in the past, that wasn't an 'issue' so I wonder if it's just a bug in i1P. Marc made them many years ago and at the time, they didn't have those few, odd patches.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #9 on: April 25, 2018, 02:11:48 pm »

Yes, but if you toggle to the print view, they disappear. And in the past, that wasn't an 'issue' so I wonder if it's just a bug in i1P. Marc made them many years ago and at the time, they didn't have those few, odd patches.
Yup, clipped to 0 or 255. I have never tried a near neutral patch set alone thinking that I1P wouldn't be able to handle it so thanks for the info Andrew. I do have a need to print high quality, precision, B&W that I use for characterizing lenses/imager registration. These have overlapping sine responses and the prints need to be very linear in range. They might benefit from a B&W specific profile now that I know I1P works in that mode.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #10 on: April 26, 2018, 01:29:38 am »

Doug, if you don't have this .pxf file, might be useful. Made by former X-rite employee Marc Levine and it contains 2500+ neutrals and near neutrals.

I tried to make a profile for the 9800 using this patch set. I1Profiler had a memory segment read fault within a second of trying to create the profile.

I then added the data from 60 coarse, non neutral, RGB patches from an earlier dataset. These were things like RGB(0,85,170)  and all other values is steps of 85.  As I said, very coarse. I1Profier promptly made a profile.

I tested it's black and white Lab response and it was very good. Should make excellent black and white profiles but one may need to add a small number of color patches to keep I1Profiler happy. Obviously it's only useful for black and white.

The 60 patches use this pattern:

0 0 85
0 0 170
0 0 255
0 85 0
0 85 85
0 85 170
etc
Logged

Rhossydd

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3369
    • http://www.paulholman.com
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #11 on: April 26, 2018, 04:06:28 am »

Interesting patch set. When loaded there are a few scattered strong colors which disappear when switching to the printable chart.  These are the result of byte wrapping since the RGB values go lower than 0, down to -3 and up to 260. They are clipped to 0 and 255 when printed.
That seems like an error to me.

Possibly worth wading through the file and removing any errors like this.

« Last Edit: April 26, 2018, 04:13:09 am by Rhossydd »
Logged

GWGill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 608
  • Author of ArgyllCMS & ArgyllPRO ColorMeter
    • ArgyllCMS
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #12 on: April 26, 2018, 04:15:32 am »

I tried to make a profile for the 9800 using this patch set. I1Profiler had a memory segment read fault within a second of trying to create the profile.
Not very surprising from my perspective. It would either need a lot of special case code, or some sort of auto-monochrome type detection to handle a dataset that is not much more than one dimensional. I'm sure ArgyllCMS would barf in some fashion as well (though hopefully not crash!). I'd imagine that similar bad things happen if you try and create a CMYK profile from a data set that has actually been squeezed through an RGB or CMY print path, and therefore only has 3 degrees of freedom....
Logged

Alan Goldhammer

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4344
    • A Goldhammer Photography
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #13 on: April 26, 2018, 08:26:33 am »

Doug,

I'm a little unclear on one aspect of your testing.  Through the work that you have done, do you end up with a single set of patches that are read for profiling or do you run a first set to get a rough profile and then refine it with a second one? 

I'm running ArgyllCMS (unwilling to pay for i1 Profiler ;D) and do a first preconditioning profile with 924 patches (2 letter size pages) and using that profile create the final one with Argyll generating neturals using the -N command with 1848 patches (4 letter size pages).  I'm not automated and read the patches with an i1-Pro.  The profiles for my Epson 3880 appear to be very good in terms of the critical colors in test prints.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #14 on: April 26, 2018, 10:37:58 am »

That seems like an error to me.

Possibly worth wading through the file and removing any errors like this.
It just means when the engineer created the patch set generating program they didn't bother clipping the extremes. To get the full range of neutrals with the easiest program you run above and below the RGB limits on some channels. Since it just looks funny in the patch set view but is clipped correctly by the I1Profiler program, no harm is done. The effect is only cosmetic and X-Rite handles (clips) these correctly.
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #15 on: April 26, 2018, 10:40:01 am »

It just means when the engineer created the patch set generating program they didn't bother clipping the extremes. To get the full range of neutrals with the easiest program you run above and below the RGB limits on some channels. Since it just looks funny in the patch set view but is clipped correctly by the I1Profiler program, no harm is done. The effect is only cosmetic and X-Rite handles (clips) these correctly.
Exactly, built in a very old version of i1P and the newer version just doesn't treat it the same 'visually'. So a new bug or an old bug, nothing unexpected from the i1P engineering team.  :'(
As I indicated, when I first got this target from Marc, these 'issues' didn't show up.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #16 on: April 26, 2018, 11:04:10 am »

Not very surprising from my perspective. It would either need a lot of special case code, or some sort of auto-monochrome type detection to handle a dataset that is not much more than one dimensional. I'm sure ArgyllCMS would barf in some fashion as well (though hopefully not crash!). I'd imagine that similar bad things happen if you try and create a CMYK profile from a data set that has actually been squeezed through an RGB or CMY print path, and therefore only has 3 degrees of freedom....
Quite!  Though I've, very infrequently, had the same fault on normal data sets. Usually when using custom illuminants. Strongly suggests ill conditioned matrixes.

Adding a very small set of boundary colors did make the program happy.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #17 on: April 26, 2018, 11:28:29 am »

Doug,

I'm a little unclear on one aspect of your testing.  Through the work that you have done, do you end up with a single set of patches that are read for profiling or do you run a first set to get a rough profile and then refine it with a second one? 

I'm running ArgyllCMS (unwilling to pay for i1 Profiler ;D) and do a first preconditioning profile with 924 patches (2 letter size pages) and using that profile create the final one with Argyll generating neturals using the -N command with 1848 patches (4 letter size pages).  I'm not automated and read the patches with an i1-Pro.  The profiles for my Epson 3880 appear to be very good in terms of the critical colors in test prints.
Except for B&W, I also do an iterative process to find additional RGB patches for both PK and MK papers. But, after that I use the same patch set as appropriate for each ink so I don't do subsequent iterations.

I also make specialized sets when optimizing B&W but these don't normally use iterative profile generation processes. Rather, I test the raw RGB device response and make patch sets from that.
Logged

narikin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1371
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #18 on: April 28, 2018, 07:27:22 pm »

Doug, if you don't have this .pxf file, might be useful. Made by former X-rite employee Marc Levine and it contains 2500+ neutrals and near neutrals.

Hah, I was planning to send that to Doug myself next week when back in US.

I still use it sometimes, no problems on my printer/profiling workstation with a 2016 version of i1 profiler.

why do you think x-rite pulled it from the 'blog' availability? It disappeared without explanation.
Logged

smilem

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 297
    • Color Management Services
Re: I1Profiler Optimized Target Set With Full Neutrals and Near Neutrals
« Reply #19 on: May 30, 2018, 05:46:28 pm »

Hah, I was planning to send that to Doug myself next week when back in US.

I still use it sometimes, no problems on my printer/profiling workstation with a 2016 version of i1 profiler.

why do you think x-rite pulled it from the 'blog' availability? It disappeared without explanation.

Maybe because X-rite are bastards to begin with?
They never invented anything, and their XRGA is because they missed few years somehow and never so they manufacture spectros not up to standard. Anyway the near neutral patch set is very good, but crashes many times. Perhaps that is why it was removed.
Logged
Pages: [1] 2   Go Up