Luminous Landscape Forum

Raw & Post Processing, Printing => Colour Management => Topic started by: rasworth on April 15, 2012, 04:22:41 pm

Title: i1Profiler test chart generation
Post by: rasworth on April 15, 2012, 04:22:41 pm
Supposedly the i1Profiler built-in chart generator is the way to go, so I cranked out a couple for one and two 8.5x11 sheets, 405 and 810 patches respectively.  My intent was to fill the letter size page with as many patches as possible readable by an i1Pro, which is easy enough to do using the built-in tool.

The single page 405 patch target contained 9 "neutral" patches, i.e. 0-0-0 to 255-255-255, equally spaced.  The two page 810 patch target contained 23 neutral patches.  I then went for the maximum number of patches to fill up a 13x19 sheet, turns out for efficiency one has to create two 9.5x13 targets and rotate and stitch together to more or less fill up the area.  I was somewhat surprised to find this resulted in only 12 neutral patches.

I realize driving the printer with equal RGB value patches may or may not actually create neutral patches on a given paper, but I was still surprised at the wide range, or more specifically going to more patches can greatly reduce the neutral inputs.  Anybody have an idea as to the basic workings of the target generation algorithm?

Richard Southworth
Title: Re: i1Profiler test chart generation
Post by: bill t. on April 15, 2012, 05:27:55 pm
Perhaps with larger numbers of samples it's easier to infer where "neutral" is, smoother curves or something.  With smaller numbers of samples perhaps interpolation is less accurate and it's necessary to be more literal about tying down the neutral points.  How's that for a WAG?
Title: Re: i1Profiler test chart generation
Post by: JeffKohn on April 15, 2012, 05:43:24 pm
Yeah and I would be more interested in how many "almost neutral" patches are included as patch count goes up. Having a lot of true neutrals like 50,50,50 in the test chart might not be as useful as having lots of patches like 50,51,50, or 50,51,49, etc.
Title: Re: i1Profiler test chart generation
Post by: Alan Goldhammer on April 15, 2012, 05:49:27 pm
Yeah and I would be more interested in how many "almost neutral" patches are included as patch count goes up. Having a lot of true neutrals like 50,50,50 in the test chart might not be as useful as having lots of patches like 50,51,50, or 50,51,49, etc.
My reading of the OP is that these are not all identical 50, 50, 50 but a B/W stepwedge and that has more steps with higher patch count.  You can do this within ArgyllCMS by specifying how many steps you want and I go for 51 which helps with smoothing.
Title: Re: i1Profiler test chart generation
Post by: digitaldog on April 15, 2012, 05:54:55 pm
Anybody have an idea as to the basic workings of the target generation algorithm?

Yup, the guys who wrote it. <G>

You can generate custom target with your own RGB values using ColorPort but unless you have specific reasons to do so (like you have company colors you want to track), just let the generator do it’s job. You want a lot of grays? Use the optimization process after initial profile creation:

http://www.i1upgrades.com/2011/08/how-to-use-the-tc-2502-gray-optimization-chart/
Title: Re: i1Profiler test chart generation
Post by: rasworth on April 16, 2012, 02:24:41 pm
Ok, cranked out an optimization on a profile I did 7 months ago for Red River Ultra Pro Satin on an Epson 3880, a semi-gloss paper with mild OBAs.  Not wanting to scan 7 pages with an i1Pro uvcut I auto generated a 405 one page optimization target, printed, scanned, and then created an optimized profile.

I noticed the 405 patch target was about 60% relatively unsaturated patches in regular patterns, seemingly intended to stay close to the neutral axis.  The remaining patches were generally very saturated, a crazy quilt all over the map.

I had generated the original profile with 1000 i1Profiler auto generated patches, maxing out the number that would fit on a 13x19 sheet with one image.  I printed test images for the old and new profiles and compared, nothing different I could see.  I examined both profiles in ColorThink, very slight difference in the neutral curves but not enough to make a visible difference.  My conclusion is for a well behaved printer (sample of one) the original profile is sufficient, no value added with optimization.

Richard Southworth
Title: Re: i1Profiler test chart generation
Post by: digitaldog on April 16, 2012, 02:31:21 pm
Not wanting to scan 7 pages with an i1Pro uvcut I auto generated a 405 one page optimization target, printed, scanned, and then created an optimized profile.

A 405 patch optimization of what? If you are having i1P just build a colored target, then your conclusions are similar to mine. And using an iSis, I did testing where the secondary optimization target had thousands of patches. More than the original. Running the profile through color patches, and then placing that into ColorThink Pro to generate a color list, the max dE’s were below 1. So in that case yes, thousands of colored patches generated for optimization was useless. Now try the 2500 patch gray target of Marc’s. I suspect you’ll see a difference as I did on my 3880. Here you are targeting just neutrals and off neutrals and a boat load of them. IOW, a 2500 patch target like Marc’s and a 2500 (heck, 5000 patch) optimization from i1Profile using other colors is a big difference in terms of the end results.
Title: Re: i1Profiler test chart generation
Post by: rasworth on April 16, 2012, 04:00:04 pm
I wasn't trying to suggest that a targeted optimization, i.e. for neutrals, wouldn't provide a benefit.  However, without other than an i1Pro I'm not willing to take on the 2500 patch target.  I read the link you provided, the poster stated he would work on a smaller target but so far not available.  The real question to me is how many patches are required to achieve meaningful results on improving neutrals, can we get the number down to something that will fit on one letter size, or 13x19, or something more manageable for us common folk.

Richard Southworth

Added by edit - Actually it's not as bad as I implied, would take two 13x19 sheets to print out, total of 93 lines to scan with an i1Pro.  Not a pleasant task but not that bad for one or two profiles.  Maybe someday.

Added by second edit - Oops, brain cramp, would require 3 13x19 sheets.
Title: Re: i1Profiler test chart generation
Post by: Erland on February 17, 2015, 08:40:30 am
The link for the massive gray chart doesn't exist any longer. Anyone who has it?
Title: Re: i1Profiler test chart generation
Post by: howardm on February 17, 2015, 09:06:52 am
you mean this one???????????

https://www.dropbox.com/s/g04qckm1k1jooqh/tc_2502_lab-gray_sco.cxf?dl=0
Title: Re: i1Profiler test chart generation
Post by: Erland on February 18, 2015, 03:55:48 am
Wow! Thanks!
Title: Re: i1Profiler test chart generation
Post by: digitaldog on February 18, 2015, 10:36:38 am
There's a smaller 900 patch target too if invested.
Title: Re: i1Profiler test chart generation
Post by: Erland on February 18, 2015, 03:40:08 pm
It is? Sound better, a bit less work. Do you have it Andrew?
Title: Re: i1Profiler test chart generation
Post by: digitaldog on February 18, 2015, 03:45:46 pm
It is? Sound better, a bit less work. Do you have it Andrew?
It's less to measure of course. Is there a benefit to this? Sometimes. YMMV. Anyway, here's the file:
 http://digitaldog.net/files/TC_900_LAB-GRAY_SCO.cxf
Title: Re: i1Profiler test chart generation
Post by: Erland on February 26, 2015, 02:06:53 am
Andrew, Thank you!
I've tried it with an optimization of my current profile and it did indeed give me a slightly better greyscale.
And right now I'm trying it with my screen calibration, since it has a slight pinkish cast on a greyscale gradient.
Title: Re: i1Profiler test chart generation
Post by: howardm on February 26, 2015, 07:48:30 am
As a possible aside, I'm not seeing a relatively 'easy' way to simply add N patches to the initial set of i1P patches vs. doing the optimization step.

The only way I've imagined so far is to save the initial/smart patch set in CGATS or something and then hack/add to it and then reload it.

Am I missing something?
Title: Re: i1Profiler test chart generation
Post by: Erland on February 27, 2015, 02:23:27 pm
I have not found any other way either. Wish they would make a test chart generator with user selectable variables.
Title: Re: i1Profiler test chart generation
Post by: digitaldog on February 27, 2015, 03:30:30 pm
I have not found any other way either. Wish they would make a test chart generator with user selectable variables.
There's ColorPort (free). 
Title: Re: i1Profiler test chart generation
Post by: howardm on February 27, 2015, 03:32:36 pm
I saw/have Colorport but I dont see how to take a i1P generated patch set and then *add* more patches to it (ie. more near-grays (assuming I have a .cxf or .ti1 for the near-grays)
Title: Re: i1Profiler test chart generation
Post by: digitaldog on February 27, 2015, 03:42:04 pm
My main suggestion for ColorPort was to build the entire patch set there. You can import a CGATs file of color values of which you can build targets or even just type in RGB Values you wish for the path(s).

If you wanted to use the actual color patch generation engine from i1P, then add more patches, you'd need ColorThink too. You would create the target as usual in i1P and save off a TIFF. Then you'd need to resample that down so 1 patch equals 1 pixel (I use Photoshop, Nearest Neighbor). Import that pixel file into CT and Extract All Colors. Save as ColorList, import into ColorPort. Add more colors to CP as you desire. But (big but), you may need to do all the measuring in CP instead of i1P as I'm not certain i1P will properly import the reference file that CP creates (it's XML). Leave it to X-rite to produce two products that share many functions and can't talk to each other via their file formats. Now if you convert the XML into a text file i1P can import, that should work.

This all might be possible using PatchTool too. It might open both CP and i1P's reference files and build it as one and I know it can export .cxf that i1P can read. Funny enough, i1P exports .cfx files it can't even reopen! But PatchTool will open it and when re-saved through PT, the .cxf then opens in i1P. Whatever the engineers at X-rite are smoking to make this dog and pony show is strong stuff.  ::)
Title: Re: i1Profiler test chart generation
Post by: howardm on February 27, 2015, 03:47:31 pm
Hmmm.  Thank you very much but I sort of lost interest in this at step #75 and purchase approx $400-500 of additional software.

I think I'll just do the optimization step if I'm so motivated.  Really wish there was a simpler way.

But you're right.  Serious crack-smoking going on out there.

I'll take a look at using CP only or maybe in conjunction w/ PT.  CT is a bridge too far  :-\
Title: Re: i1Profiler test chart generation
Post by: Erland on March 27, 2015, 04:58:14 am
You can try playing with the chart generation in i1Profilers Optimization module. I create a chart with spotcolors, and an black and white image with 30 patches, and 1300 random patches etc. Creates very good charts with a lot of neutral and near neutral patches.
Title: Re: i1Profiler test chart generation
Post by: hugowolf on March 28, 2015, 07:41:45 pm
I have been experimenting with the profile optimization process in iProfile (iPublish). I have been comparing profiles made with 2, 4, and 6 US letter sized sheets (680 patches on each), with profiles made with fewer sheets on the first iteration, and then a page of optimization patches.

I get better profiles, more accurate and better linearization, with 2 sheets and 1 automatically generated optimization sheet, than with 4 sheets once off. With six sheets being too close to call against 2 plus 1.

It adds a day to profile creation, but saves paper and ink.

Brian
Title: Re: i1Profiler test chart generation
Post by: howardm on April 01, 2015, 08:17:23 am
Thanks Brian, that's good paper-saving info :)

How many optimization patches usually get generated automatically by the s/w?  And then do you 'supplement' ?
Title: Re: i1Profiler test chart generation
Post by: hugowolf on April 01, 2015, 02:02:45 pm
Thanks Brian, that's good paper-saving info :)

How many optimization patches usually get generated automatically by the s/w?  And then do you 'supplement' ?

I just let it automatically generate the all optimization patches from the profile made from the first two sheets. I have yet to try adding an image or importing specific patches, or any other combination of the three available options. A sheet will take slightly more than 680 patches, but going to 681 I lost some patches that I wanted to keep - I think some light off neutrals. It is like when you generate the initial patch set, and even though there may be room for more on the sheet, adding a few more can make a whole group of neutrals disappear.

To get 680 on one sheet, I have to reduce three of the four borders to zero in iProfiler. I loose the row identification numbers, everything else fits on. One sheet I fed was actually only 7.5 x 11, and the first row of target patches were printed almost full bleed and were only 6mm instead of 9mm, but they read ok.

This is on an 9890. which has minimum borders of 3mm top, right, and left, and a trailing edge minimum of 0.56 inches for sheets. The 9890 only does borderless for 10, 13, 16, 17, 24, 36, and 44 inches, and rolls only not sheets of any size. You might get more patches from your 3800, which you can set to print borderless, even though you are not inking to the edge. iProfiler doesn't seem to communicate to well with the printer as far as borders go.
 
I have only used the technique for three papers so far, all of them matte fine art: Canson Rag Photographique 310, 210, and Hahnemühle German Etching.

Brian A