Pages: [1]   Go Down

Author Topic: Epson 850 profiling to a printed targets instead of IT8 charts?  (Read 1792 times)

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197

I picked up the Epson 850. Thanks to Mark Segal for his excellent review and paper. Saved a lot of time tracking down the reference data for the Xrite IT8 chart and general process flow.

However, the chart the scanner came with, which has a date of 6/17 has a white point measurement that's a bit over 2 dE off compared to a spot check with an I1. in the a*/b*. Also, a white sheet of neutral, non OBA, paper should be placed on top of the chart when scanning it. The chart is relatively thin and there is 1.3 increase in L* when measured with the white sheet on top. This corresponds to the standard way ICC profiles are made while the graphic arts uses a black background. The scanner will measure based on a black background as this scanner does not press down with a white pad.

To check the profile from the IT8, I scanned a printer target and compared the dE00 values for all in gamut colors. They had a pretty close baseline dE00 of just under 3. Considering that the IT8 chart is made on actual photo paper and is presumably best for producing profiles to scan other photopaper prints, this seems a reasonable result as it isn't measuring the chart spectrally.

However, I also want to make profiles for scanning my prints. The most obvious way to do this is with Argyll. Alternately, I suppose I could make a printer profile chart in IT8 format then scan it and put it into the format expected by I1Profiler's scanner module. Doable, but Graeme's Argyll is a bit more flexible so I'll try that first.

Ideally, I'd like to get the scanner reading colors as close as possible to what the iSiS produce so I can compare apples to apples.

Anyone have experience doing something like this?

Edit: I found the white backing pad packaged with the slide/film holders so no white backing needed.
« Last Edit: May 18, 2018, 02:45:11 pm by Doug Gray »
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Epson 850 profiling to a printed targets instead of IT8 charts?
« Reply #1 on: May 18, 2018, 02:08:54 am »

After trying the I1Profiler scanner module I switched to the Argyll 2.0.0 stuff. Works great. The scanner has some areas that are OOG but the averages are quite good. Ave dE is 1.2 and max is 6 with the XYZ PCS lut selection. With the standard LAB PCS Ave dE goes up to 1.6 and 9. For printed CC colors it's well under 1. The profile is based on printer inks. Next to see how far off scanning a regular CC card would be. Pretty easy to get running.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Epson 850 profiling to a printed targets instead of IT8 charts?
« Reply #2 on: May 19, 2018, 12:19:45 am »

First, creating input profiles with the 850 using printed targets and Argyll software is fairly straightforward. Average dE00 is around 1 but increases in the darker patches. The Epson s/w makes it pretty easy to select the input profiles when you scan. However, when you do they automatically convert to only a handful of RGB working spaces, notably sRGB and Adobe RGB. If you want a wider space like ProPhoto, you need to scan w/o color management then attach the input profile to the image file when opening it in Photoshop. Then convert to ProPhoto.

interesting error in the Epson 850. The average brightness and color of a sheet  that you are scanning impacts the values of the patches read. For instance, if you print a patch that is L*=90 alone on a white sheet, the patch, as well as the rest of the white paper, will read with a higher brightness compared to scanning a darker sheet such as a typical print with an average L* around 50 (18%). This suggests that light reflecting off the sheet is bouncing off the scanner insides and increasing the overall luminance*. The effect is similar to camera glare where one shoots a black patch against a black background and compares it to the same patch against a white background. The former has a slightly higher L*.

* The scanner has a two rows of white leds on either side of the scanning sensor. They are arranged so the leds interleave between the two sides. So the effects of adjacent reflected light changing readings should be more impacted by horizontal colors than differences along the vertical.

That said, by profiling the scanner to a printed target, scans of prints made on the same printer are quite good. For instance, I scanned a print of colorchecker patches and the ave dE00 was 1.1 with a max of 1.8. The largest errors were in the darker patches which may be due to light reflected off the glossy surface. OTOH, scanning an actual colorchecker produced significantly larger errors, about 2.5 dE00. This may be a combination of factors such as the matte instead of glossy surface and especially spectral differences even though the patches measure Lab values closely with a spectro. I did not see any significant difference using the IT8 chart compared to using my printed target when scanning an actual colorchecker. In fact the IT8 chart produced slightly worse results.

But all that aside, the scanner does produce quite smooth transitions and early testing indicates it may work very well for identifying how smooth gradients print.
« Last Edit: May 19, 2018, 12:31:21 am by Doug Gray »
Logged

BrianToth

  • Newbie
  • *
  • Offline Offline
  • Posts: 43
Re: Epson 850 profiling to a printed targets instead of IT8 charts?
« Reply #3 on: May 24, 2018, 12:16:52 am »

After you helped me sort out my printing issues (http://forum.luminous-landscape.com/index.php?topic=124335.80) I got back to scanning and I'm pretty much back to square one with that. Like I mentioned in that thread, I came up with a set of scripts to do some systematic analysis of every profile I could create. Happy to share, but I have a feeling you have your own system in place that's more automated. :)

1. I took a ColorChecker that I printed and measured with my spectrophotometer.

2. Scanned the ColorChecker and my three targets (IT8, CC24, DCCSG) using Epson Scan w/ no color management, and with Gamma 2.2 but a reset histogram; VueScan RAW 48bit, RAW 24bit, and with Color None and DeviceRGB as output.

3. Wrote a script to generate every combo of Argyll profiles from the scans and used i1Studio where it would let me (it didn't like the RAW images).

4. Recorded the RGB of the corresponding test ColorChecker print for each scan method and wrote another script to take my ICCs and run the RGB through them using Argyll's icclu tool.

5. Then I took those Lab numbers and wrote another program in Python to output the ∆E(2000) between the resulting Lab values and the Lab of the original print.

I've attached a sample of my ∆E for Epson Scan and a few different profiles vs. my printed ColorChecker if you're curious. (The image with the patches have Lab rounded to whole numbers unfortunately, but it's still pretty representative of my issues even with other photos.)  I didn't notice significant difference between the different scan methods. In some cases the RAW worked better, in others a more normal scan was preferable. My custom target I made with Argyll on the same paper yielded the best results for the ColorChecker I printed… but didn't work at all for my photo collection I'm trying to scan.

I'd be curious to see if you have luck getting a general profile that's good enough or if you're only satisfied with custom profiles for each paper you're trying to reproduce.
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
Re: Epson 850 profiling to a printed targets instead of IT8 charts?
« Reply #4 on: May 24, 2018, 07:37:30 am »

I'm doing a considerable amount of work on scanner profiling that I shall be writing-up within the next few weeks. For the moment, there is only one point you made I would like to respond to, and that is at the end where you talk about paper. I'm not sure what you means by that, but if by "paper" you have "prints" in mind, I should just mention for now that a scanner profile is an input profile to characterize the response of the input device (the scanner). A paper profile is an output profile to characterize the response of the printer/paper combination you will use to make the final product; there are many choices here as we know. The general principle is that you rely on the scanner profile to help produce an acceptable digital file of the media being scanned (i.e. the original source material) and you rely on the printer/paper profile to help produce an acceptable print of the file created from the scanner. The scan should be amenable to multi-purposing, and the purpose of the profile is to characterize that one device.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Epson 850 profiling to a printed targets instead of IT8 charts?
« Reply #5 on: May 24, 2018, 10:58:15 am »

I'm doing a considerable amount of work on scanner profiling that I shall be writing-up within the next few weeks. For the moment, there is only one point you made I would like to respond to, and that is at the end where you talk about paper. I'm not sure what you means by that, but if by "paper" you have "prints" in mind, I should just mention for now that a scanner profile is an input profile to characterize the response of the input device (the scanner). A paper profile is an output profile to characterize the response of the printer/paper combination you will use to make the final product; there are many choices here as we know. The general principle is that you rely on the scanner profile to help produce an acceptable digital file of the media being scanned (i.e. the original source material) and you rely on the printer/paper profile to help produce an acceptable print of the file created from the scanner. The scan should be amenable to multi-purposing, and the purpose of the profile is to characterize that one device.
Hi Mark,

I'm looking forward to your scanner profile results.

Brian is referring to profiling a scanner with a printed target (reflectance profiling only) instead of the usual Colorchecker, IT8, etc. I had brought this up earlier and he is discussing his various test results.

Profiling a scanner with a printed target that is also spectro scannable has these advantages:

  • Use of the same instrument for reading Lab values. There is considerable inter-instrument variation. This can make replication accuracy verification more difficult.
  • Allows easy and fast creation of different patch sets.
  • Optimal matching of generated scanner profile to the print's ink set. CC's, IT8's have different spectra and L/I becomes more of a factor.
Disadvantages:
  • Likely less useful for scanning chem. lab processed photos. IT8 based profiles should be closer.
  • Argyll workflow not familiar to many.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Epson 850 profiling to a printed targets instead of IT8 charts?
« Reply #6 on: May 24, 2018, 11:02:12 am »

I'd be curious to see if you have luck getting a general profile that's good enough or if you're only satisfied with custom profiles for each paper you're trying to reproduce.
I don't believe there is significant differences between scanner profiles for different printed papers and suspect it's mostly ink based and that paper variation makes little difference. However, matte v glossy types may benefit. I haven't any data regarding whether this will matter or not.
Logged

BrianToth

  • Newbie
  • *
  • Offline Offline
  • Posts: 43
Re: Epson 850 profiling to a printed targets instead of IT8 charts?
« Reply #7 on: May 24, 2018, 05:03:28 pm »

Brian is referring to profiling a scanner with a printed target (reflectance profiling only) instead of the usual Colorchecker, IT8, etc. I had brought this up earlier and he is discussing his various test results.

I don't believe there is significant differences between scanner profiles for different printed papers and suspect it's mostly ink based and that paper variation makes little difference. However, matte v glossy types may benefit. I haven't any data regarding whether this will matter or not.

That's what I meant, sorry for the confusion. The details are important in this forum. :) Yes, making a target using the same inks that the test "photo" was printed with yielded the best reproduction results… which was not surprising to me. My original hope was, naively, that simply having a lot more color patches to sample would produce a better overall profile, but that is not necessarily the case for reasons that I now understand. 

I also recently took the time to make my own reference files for my two ColorCheckers, which had some noticeable differences from the canned reference files from X-Rite. I hopped that having a more accurate reference, along with the spectral data (that X-Rite doesn't include in their references files) would help nudge the results in the right direction, but I couldn't tell any difference at all, at least with the samples I was testing with.

I would expect the IT8 to work the best, given that's what it was designed for, but I can't seem to overcome the blue casts that it creates.

I look forward to reading what Mark publishes as well.
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
Re: Epson 850 profiling to a printed targets instead of IT8 charts?
« Reply #8 on: May 24, 2018, 08:08:58 pm »

That's what I meant, sorry for the confusion. The details are important in this forum. :) ...........

Well, yes - details are important to the extent they help clarify what one is talking about  :-). Thanks Brian and Doug for the clarification; I now see what that comment referred to - totally different from what I initially thought, so my immediate previous post is moot.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Epson 850 profiling to a printed targets instead of IT8 charts?
« Reply #9 on: May 24, 2018, 08:53:33 pm »

Well, yes - details are important to the extent they help clarify what one is talking about  :-). Thanks Brian and Doug for the clarification; I now see what that comment referred to - totally different from what I initially thought, so my immediate previous post is moot.

I've progressed to the point that I'm able to reduce the variance between random black and white patches by a very large factor. The good  news is that I believe the algorithm can be applied to any scan image to essentially eliminate this effect in any scan. It should also be possible as well to apply it to the IT8 targets which would be really good as that would provide improved profiles for scanning regular photos, not just prints. We'll see.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Epson 850 profiling to a printed targets instead of IT8 charts?
« Reply #10 on: May 25, 2018, 10:35:08 pm »

I've now verified that my algorithm also reduces reflectance induced variance of random color patches as predicted by the model. The residual error also has a significant, though now relatively small, luminance component. However, that component should be stationary so it can also be corrected by scaling against a completely unprinted paper scan. After that I'll take a look at the significant variation amongst darker patches by scanning an all black printed sheet.
Logged
Pages: [1]   Go Up