Pages: 1 [2]   Go Down

Author Topic: i1Profiler smoothness control effect?  (Read 2567 times)

rasworth

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 473
Re: i1Profiler smoothness control effect?
« Reply #20 on: April 26, 2019, 03:23:43 pm »

Doug, I understand the intent was for final BW profiles, I was speculating on its usage as a general purpose training profile also.

Thanks again,

Richard Southworth
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: i1Profiler smoothness control effect?
« Reply #21 on: April 28, 2019, 01:01:15 am »

Update:

I've done a lot of testing focused on getting the right recipe for making good tracking target patches. What works very well is a combination of full coverage, 7 (instead of the original 3) patches per step, together with a closer spacing of neutrals only, offset to the device neutral axis as determined by the initial profile.

The main breakthrough was creating, printing, and scanning multiple test sets with different spacings by combining them into one larger CGATs file which is then randomized so all the test targets are randomly spread across the printer pages. This spreads the printer drift, which is significant from one page to the next so it affects all the targets. The pages are then scanned, de-randomized, and each "test chart" is pulled out into separate CGATs files which are then loaded into I1Profiler and profiles made.

Then, each profile is used to create, in device space based on each of the tested profiles from above, known colors which are the 18 colored CC values, as was as neutrals from L*=3 to L*=93 in steps of 3. These are also randomized, printed, and de-randomized.

These approaches provide excellent, consistent results and allow good comparison between RGB patch set performance. The large number of neutrals is especially good for tuning neutrals. And this is where things are most critical because visual sensitivity to small changes is highest in and around neutrals.

I am very happy with the results. None of the profiles exhibit the anomalies of my initial tracking profile design.

All my testing was being done in Matlab and I will now port the changes to C++ so it can be made available to all interested.
Logged
Pages: 1 [2]   Go Up