Pages: [1]   Go Down

Author Topic: Updated link for Bill Atkinson's "28 Balls" test-print (& curious how to eval)  (Read 16707 times)

l_d_allan

  • Full Member
  • ***
  • Offline Offline
  • Posts: 208
    • Berean photos

I've noticed several references to the "28 Balls" test-print by Bill Atkinson, but I've had trouble locating a current link. Here is a current link, based on correspondence with BA (who graciously gave permission to provide this link ... thx):
http://www.dropbox.com/sh/ka35g3gtyd10823/iRGJexS5hA/Twenty-Eight%20Balls.tif.zip

Also the link for "14 Balls", which is a subset:
https://www.dropbox.com/sh/ka35g3gtyd10823/-FRLLos-8N/Fourteen%20Balls.tif

Note that you may have to click several times on the "Download" button to get the download started. It may or may not work ok with https and/or http.

I'm curious how people who have experience with "28 Balls" use the image file to evaluate prints. My understanding is that "28 Balls" is a rather extreme synthetic test-print to challenge a printer + profile.

There is an essay that accompanies the OutbackPhoto.net test print (based on a test-print from BA?), that explains how to use it, and what to look for. I'm unclear about "what to look for" with the 28 Balls test print. The email from BA had some comments, but this CM semi-newbie is fuzzy on how to comprehend them.
Logged
retired in Colorado Springs, CO, USA ... hobby'ist with mostly Canon gear

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/

I'm curious how people who have experience with "28 Balls" use the image file to evaluate prints. My understanding is that "28 Balls" is a rather extreme synthetic test-print to challenge a printer + profile.

It really puts a profiler's color engine to a good test. For example, I've been comparing X-Rite's i1Profiler to ColorLogic's Copra. This one page shows a lot! I've discovered where i1P does a better job mapping out of gamut blues compared to Copra where blues shift a tad magenta. Otherwise Copra puts i1P's engine to task! By providing these images to ColorLogic, I hope they can fix that small area of color space. When dealing with OOG colors, a profile engine can weigh towards hue or lightness and you see these decisions in Bill's image really well. I'm referring here to the perceptual table but you can use it with smaller visible differences with other rendering intents. Where one profiler shifts hue, the other lightness.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

l_d_allan

  • Full Member
  • ***
  • Offline Offline
  • Posts: 208
    • Berean photos

I'm not sure I really understand "Rendering Intents".

My impression is that "28 Balls" with very saturated, synthetic colors might tend to print better with "Perceptual Intent" rather than "Relative Color". My speculation is that "Perceptual" might "spread out" the very saturated colors, so there might be less banding. My further speculation is that "Relative Color" might group more of the very saturated colors into the same color, and might result in more banding.

Or not?
Logged
retired in Colorado Springs, CO, USA ... hobby'ist with mostly Canon gear

xpatUSA

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 390
    • Blog

I'm not sure I really understand "Rendering Intents".

My impression is that "28 Balls" with very saturated, synthetic colors might tend to print better with "Perceptual Intent" rather than "Relative Color". My speculation is that "Perceptual" might "spread out" the very saturated colors, so there might be less banding. My further speculation is that "Relative Color" might group more of the very saturated colors into the same color, and might result in more banding.
Or not?

Not sure that rendering intent is a factor in banding, but this tutorial is one of the better ones in terms of understandability:

http://www.cambridgeincolour.com/tutorials/color-space-conversion.htm

OT but often, when the profile says "perceptual" you don't actually get a perceptual conversion or rendition. Something to do with simple matrix transformation.
Logged
best regards,

Ted

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/

I'm not sure I really understand "Rendering Intents".
http://www.ppmag.com/reviews/200405_rodneycm.pdf

Quote
OT but often, when the profile says "perceptual" you don't actually get a perceptual conversion or rendition. Something to do with simple matrix transformation.
Version 2 matrix profiles only have the colorimetric table. So in say Photoshop, you can select a conversion from Adobe RGB to sRGB and select Perceptual, you'll get a Relative Colorimetric rendering. There's no Perpetual table. If you examine these simple ICC profiles, they are pretty small (4K or so). Output profiles (LUTs) are much larger due in part to having these additional tables.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

l_d_allan

  • Full Member
  • ***
  • Offline Offline
  • Posts: 208
    • Berean photos

Not sure that rendering intent is a factor in banding,

I was thinking you could get some banding where the color gradient values became out-of-gamut for the specific printer, causing multiple RGB values to be compressed/mapped to one RGB value that the printer could handle. Or not?

Quote
but this tutorial is one of the better ones in terms of understandability:
http://www.cambridgeincolour.com/tutorials/color-space-conversion.htm

Thanks. Very helpful.

At the conclusion of the Cambridge-In-Color tutorial, there was the paragraph:
Quote
The decision about when to use each of these depends on image content and the intended purpose. Images with intense colors (such as bright sunsets or well-lit floral arrangements) will preserve more of their color gradation in extreme colors using perceptual intent. On the other hand, this may come at the expense of compressing or dulling more moderate colors. Images with more subtle tones (such as some portraits) often stand to benefit more from the increased accuracy of relative colorimetric (assuming no colors are placed within the gamut mismatch region). Perceptual intent is overall the safest bet for general and batch use, unless you know specifics about each image.

My "take-away" from the above was that ...
  • With images that don't have all that much overall gamut, the choice of rendering intent for printing probably doesn't matter much, but Perceptual a good choice. Or not?
  • With portraits that also had lots of out-of-gamut colors (such as the Adobe test print Colorfile_AdobeRgb.tif with lady (Shirley?) in very colorful fruit-hat), to preserve the accuracy of flesh tones, the preferred rendering intent would typically be "Relative Colorimetric". Or not?
  • With lots of out-of-gamut colors such as scenery, and without flesh tone accuracy being a high priority, then the preferred rendering intent would typically be "Perceptual". Or not?
« Last Edit: July 13, 2013, 08:10:30 pm by l_d_allan »
Logged
retired in Colorado Springs, CO, USA ... hobby'ist with mostly Canon gear

l_d_allan

  • Full Member
  • ***
  • Offline Offline
  • Posts: 208
    • Berean photos

http://www.ppmag.com/reviews/200405_rodneycm.pdf

Also very helpful. I was able to find and download 35 articles written by Andrew on color management from January, 2004 thru Oct, 2007. I'm working my way through them, and learning quite a bit. Thanks!
Logged
retired in Colorado Springs, CO, USA ... hobby'ist with mostly Canon gear

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/


My "take-away" from the above was that ...
  • With images that don't have all that much overall gamut, the choice of rendering intent for printing probably doesn't matter much, but Perceptual a good choice. Or not?
  • With portraits that also had lots of out-of-gamut colors (such as the Adobe test print Colorfile_AdobeRgb.tif with lady (Shirley?) in very colorful fruit-hat), to preserve the accuracy of flesh tones, the preferred rendering intent would typically be "Relative Colorimetric". Or not?
  • With lots of out-of-gamut colors such as scenery, and without flesh tone accuracy being a high priority, then the preferred rendering intent would typically be "Perceptual". Or not?

ICC profiles don't know squat about image content, color in context etc. They only understand a single color value of a single pixel and how it should be mapped to another value. So soft proof! Select each rendering intent (certainly Perceptual and RelCol), look at the image and pick the one you visually prefer.

Note too that the Perceptual rendering mapping follows no rules. That is, a profile for say an Epson 4900 built using profile product A will result in a different perceptual rendering than product B. The people making the decisions hope they are producing results someone will visually prefer. Much like Kodak wanted one transparency film (Ektachrome) to produce a rendering differently than Fuji (Velvia). There's no right answer. When dealing with Out Of Gamut (OOG) colors using a perceptual intent, they have to weight saturation against lightness and everyone does it differently. Recently discussed here along with Bill's Balls image (Copra vs. I1Profiler).
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

texshooter

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575

For giggles I printed this chart on Moab Entrada Bright with both Moab's ICC as well as a Colormunki ICC. With the Colormunki, the bottom right blue ball was printed pure black. But for all the other colors, the Colormunki did a much better job as smoothing the banding problem. I used perceptual rendering intent for both. Don't know what to make of it but it is what it is.
Logged
Pages: [1]   Go Up