Hi All,
I am very new to color management, so do let me know, if I should post my questions elsewhere and/or differently.
I am trying to obtain a basic understanding of the typical quality of canned profiles (mainly using Hahnemühle papers) and those I am creating myself.
Although none of the apps I tried (X-Rite or Datacolor) provide any functionality to support this use case, and none of the many forums I checked provide any instructions, I do find this useful, in order to decide, how good a canned profile is and if it is worth the resources to create a custom one, to check if my profiles/setup are up to typical quality standards (nothing seriously "broken"), if/how my setup is degrading over time, or to better understand how much effort (patches, ink, paper) is justifiable to create my own profiles.
I am very eager to learn, what I might be missing in trying to pursue such tests, as neither the developers of the major calibration tools nor anybody in the many forums I checked appear to find this useful. I am aware that average statistical errors don't tell you much about the strengths, weaknesses, and visual appearance of a profile to the human eye. So I am also looking at test images to check how the profiles perform on real images. Still, having some absolute and objective metrics in addition should be very helpful to keep tabs.
I have come up with some basic tests and would like to get some opinions on the approach and results, before I head off spending much more paper and ink.
I am using a Canon Pro-1000, i1Pro3+, i1Profiler App, PhotoShop, and the BabelColor Patch Tool.
My approach to measuring the quality of profiles:
- Choose a list of patches with known Lab values. For the below tests I took the DigitalColorChecker_M0 set of 140 patches that comes with the PatchTool, mainly because this fits on one A4 page with the i1Pro3+, the colors should virtually all be within gamut, and I could not find any other representative set of similar size
- Save a 16bit TIFF target image in Lab space out of PatchTool
- Load the image in PhotoShop and double check that the Lab values still match the original ones shown in PatchTool
- Print target image from PhotoShop with absolute colorimetric intent
- Scan print with i1Pro3+ from PatchTool
- Run the comparison function in PatchTool to obtain the deltaE* errors for each patch and some additional statistics
Following this process, I tested the Hahnemühle original canned profile for their Photo Luster paper, as well as a simple initial one I created as a "baseline" with just 293 patches.
See the picture below for the results. The left side shows the results for the canned profile, the right side for my customer one. I am a bit surprised of the size of the errors:
For the canned Hahnemühle profile, more than 101 of the 140 patches have a dE higher than 4, half of them higher than 5.5, and the worst 14 patches an error of more than 29.
For my simple custom profile it looks considerably better, but still not great. 50% of patches are above dE of 2 and the worst 10% are over 5.9. Maybe this is just due to 293 patches not being sufficient.
Thanks in advance for sharing your experience on what good really looks like!
Markus