Pages: [1]   Go Down

Author Topic: Profiling with 918 Patches  (Read 3358 times)

Mike Sellers

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 666
    • Mike Sellers Photography
Profiling with 918 Patches
« on: December 25, 2014, 01:12:49 pm »

What would be the advantage of using the advanced profiling software for my Z3100? Doesn`t this allow for printing 918 patches instead of the standard 450 or so? Would my color prints look better somehow?
Mike
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Profiling with 918 Patches
« Reply #1 on: December 25, 2014, 01:44:46 pm »

In theory yes. But there’s a point of diminishing return. 918 isn’t that many patches. I’d give it a try. In a perfect world, you’d measure 16.7 million patches. That take forever and you’d end up with a profile larger than many images. So the idea is anything less, the software has to extrapolate all the colors otherwise measured and again,there is a point of diminishing return. I usually work with a 1728 patch target because it’s fast to measure in my auto Spectrophotometer. I’ve done as many as 5000 patches. I’ve seen 100 patches then an optimization of 100 (ColorMunki) produce very good profiles. Bottom line: If you can measure the larger number of patches easily, no harm done. Might try one each way and run a few images to see if in the future there’s anything to be gained with the larger patch set.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Mike Sellers

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 666
    • Mike Sellers Photography
Re: Profiling with 918 Patches
« Reply #2 on: December 26, 2014, 10:25:33 pm »

Dog,
Did the 5000 patch profile produce prints that were superior in any way?
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Profiling with 918 Patches
« Reply #3 on: December 27, 2014, 12:53:56 pm »

Dog,
Did the 5000 patch profile produce prints that were superior in any way?
Cannot say with certainty as I’d have had to output lots and lots of files and carefully evaluate. Depends on the profile package, how well it creates those patches. Depends on how the patches are generated, depends on the instrument. Depends on the kind of output device. There can be a huge difference in quality building a profile for a press versus a well behaved device like an Epson. There’s a big difference in how many targets you measure and average, how you spread the patches across a page to handle variations in printing across that page (I often make a target that is rotated 180 degrees and ‘normal’ then average a lot of measurement data.

I recall when i1Profile first came out, I built a custom target there and in ProfileMaker Pro using the same number and did see a very sight improvement in the newer patch generation but it’s very, very subtle. For me, about 1700 patches is the sweet spot with an iSis. For a well behaved RGB device like an Epson.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Paris1968

  • Newbie
  • *
  • Offline Offline
  • Posts: 24
Re: Profiling with 918 Patches
« Reply #4 on: December 27, 2014, 01:47:13 pm »

I agree.  I use i1Profiler and an iSis, and 1700 to 1900 patches is the sweet spot.  Strangely, I get better profiles if the patches are an odd number. Although the i1Profiler can go beyond 2700 patches, I have yet to see any advantage in exceeding 1900.
Logged

Rhossydd

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3369
    • http://www.paulholman.com
Re: Profiling with 918 Patches
« Reply #5 on: December 27, 2014, 02:26:30 pm »

Strangely, I get better profiles if the patches are an odd number.
This has been discussed here before a year or two back. One key factor seems to be getting the maximum number of grey squares. A great example is going between 2033 and 2034 patches, at 2033 there are lots, add one patch more and all the monochrome patches disappear.
Logged

John Caldwell

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 704
Re: Profiling with 918 Patches
« Reply #6 on: December 27, 2014, 08:05:40 pm »

...A great example is going between 2033 and 2034 patches, at 2033 there are lots, add one patch more and all the monochrome patches disappear.


Asking as a genuine and true novice: What is it about the addition of one patch that makes the monochrome patches disappear?

Thanks you,

John Caldwell
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Profiling with 918 Patches
« Reply #7 on: December 27, 2014, 08:07:09 pm »

I’d consider it a bug.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Jim Kasson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2370
    • The Last Word
Re: Profiling with 918 Patches
« Reply #8 on: December 27, 2014, 08:38:33 pm »

In a perfect world, you’d measure 16.7 million patches. That take forever and you’d end up with a profile larger than many images. So the idea is anything less, the software has to extrapolate all the colors otherwise measured and again,there is a point of diminishing return.


I've had good results doing a profile based on the colors in the image to be printed, but it's not something you want to do for every image (PITA), and not every profiler supports it. It's something to try for that special image that just doesn't look right coming out of the printer with your standard profile. Another drawback is that you need to do some profile management if you don't want your drop-down lists to get cluttered.

Jim

PS. Andrew, I think you mean "interpolate".

John Caldwell

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 704
Re: Profiling with 918 Patches
« Reply #9 on: December 27, 2014, 08:41:37 pm »

I'll submit: Both extrapolate and interpolate could be accurate, depending upon where the data point is that you want In relation to known data points.
John Caldwell
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Profiling with 918 Patches
« Reply #10 on: December 27, 2014, 08:42:48 pm »

I think I mean extrapolate in that there are assumptions made. It might actually be interpolation but I am not sure it is as as accurate as that term (under my understanding of the two) imply.

Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Jim Kasson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2370
    • The Last Word
Re: Profiling with 918 Patches
« Reply #11 on: December 27, 2014, 09:57:00 pm »

I'll submit: Both extrapolate and interpolate could be accurate, depending upon where the data point is that you want In relation to known data points.

All the coarse patch arrays that I've created -- or seen, for that matter -- left out intermediate values, but never the extreme ones. I don't understand the rationale for doing otherwise, in the lack of a priori knowledge of the non-linearities of the 3D LUT.  

Jim

« Last Edit: December 27, 2014, 09:59:50 pm by Jim Kasson »
Logged

Jim Kasson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2370
    • The Last Word
Re: Profiling with 918 Patches
« Reply #12 on: December 27, 2014, 10:14:13 pm »

I think I mean extrapolate in that there are assumptions made. It might actually be interpolation but I am not sure it is as as accurate as that term (under my understanding of the two) imply.

Here is a set of criteria for the setting up and use of a 3D LUT from the abstract for a paper I wrote a long time ago:

"An interpolation algorithm design is divided into three parts: packing (filling the space of the input variable with sample points), extraction (selecting from the constellation of sample points those appropriate to the interpolation of a specific input point), and calculation (using the extracted values and the input point to determine the interpolated approximation to the output point). Seven principles govern the design of linear interpolation algorithms: 1) Each sample point should be used as a vertex of as many polyhedra as possible; 2) the polyhedra should completely fill the space; 3) polyhedra that share any part of a face must share the entire face; 4) the polyhedra used should have the fewest vertices possible; 5) polyhedra should be small; 6) in the absence of information about curvature anisotropy, polyhedra should be close to regular in shape; and 7) polyhedra should be of similar size."

http://spie.org/Publications/Journal/10.1117/12.208656

Going from the sampled CIE colors obtained from a grid in printer colorant space needs to be done to construct the LUT. You always want to sample the extremes in printer colorant space. In a CMY printer, you always want 0/0/0, 0/0/1, 0/1/0, 0/1/1, 1/0/0, 1/0/1, 1/1/0, 1/1/1 (assuming 1 is full scale) in your sample set.

If you meant, as I think you're saying, extrapolation in the metaphorical sense, then I won't argue with you.

Jim

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Profiling with 918 Patches
« Reply #13 on: December 27, 2014, 11:03:53 pm »

So what (which) is i1P’s patch generator doing?
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Rhossydd

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3369
    • http://www.paulholman.com
Re: Profiling with 918 Patches
« Reply #14 on: December 28, 2014, 04:40:06 am »

Asking as a genuine and true novice: What is it about the addition of one patch that makes the monochrome patches disappear?
As Andrew has said; Almost certainly a bug. Curious that they haven't fixed it in any of the maintenance builds, but I suspect it doesn't make a huge difference in result, but it looks wrong when it happens. It happens in exactly the same way at other specific numbers of patches too.
Logged

Alan Goldhammer

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4344
    • A Goldhammer Photography
Re: Profiling with 918 Patches
« Reply #15 on: December 28, 2014, 09:10:56 am »

I use 1848 patches with ArgyllCMS that include a 51 step B/W set.  This prints out on exactly four letter size pages and gives me excellent profiles.  I've tried going higher and don't see much if any difference.  I read the patch set manually with an i1Pro (don't do enough profiling to warrant buying an automated reader system).  It takes about 15 minutes to read a full set of patches which isn't too bad.
Logged
Pages: [1]   Go Up