1

I'm collaborating with another person on scanning anime cels. But first we need to confirm that our scanners produce identical colors. So we picked one paint color that we both have and compared our scans.

I'm using an Epson Perfection v600 scanner, while my partner uses an Epson Expression 10000XL. We both calibrated our scanner using the Silverfast scanning software 8.8 with a 864 patches IT8 target. And our average deltaE are 0.6 and 1.0, which mean our scans should be color accurate.

However our testing scan on the same paint color looks noticeable different. Our hues are 243 to 233, that's enough difference to tell something is wrong. But I couldn't figure out where the problem came from.

Update:

For more testing, we both make a scan of our IT8 target under our calibrated profiles and saved under ProPhoto RGB color profile. When checked for proofing in photoshop, the gamut of my scan was within the AdobeRGB range but my partner's scan, many of the patches exceed AdobeRGB.

I uploaded my scanned target here. Target 1 is mine, target 2 is my partner's.

https://i.stack.imgur.com/Pp1os.jpg

reddy
  • 315
  • 1
  • 8
  • 1
    Calibration does not give a scanner the ability to sense a color it is not capable of sensing, it only insures that the output is not further from the actual color than is absolutely necessary. The two different scanners likely have different "holes" in their coverage. It could be as simple as the difference in CRI between the respective lights that illuminate the object being scanned by each scanner. – Michael C Jan 09 '22 at 14:28
  • 1
    I'm not sure if we have enough information to be helpful. Would you be able to post pictures of both the scanned targets from both scanners and also the scanned anime cells? In addition given that you are scanning different pictures could it just be that the pictures are different? I think for a fair comparison you should scan the exact same picture on both of the scanners. – The Movie Man Jan 09 '22 at 17:53
  • I uploaded both of our scanned target, converted to sRGB. – reddy Jan 09 '22 at 18:06
  • You saved the scans as ProPhoto then uploaded as sRGB… there are 2 places for error already. Why not save them with the scanner's profile? That, of course, is why you profiled the scanners in the first place. – Tetsujin Jan 09 '22 at 19:51
  • @Tetsujin I updated the links, I thought that the image upload site doesn't support ProPhoto images – reddy Jan 09 '22 at 22:49
  • Even in ProPhoto, the two scans look noticeably different. I still don't know why you're not using the scanner profiles. – Tetsujin Jan 10 '22 at 08:59
  • @Tetsujin what do you mean by scanner profile? The images were scanned from our scanners, and our scanners were calibrated from the corresponding targets. The calibration created the icc profile for the scanners, if that's what you mean. – reddy Jan 10 '22 at 14:30
  • You have profiles for the scanners yet you don't appear to be using them. See https://www.silverfast.com/download/docu/ch7colourmanagement,completechapter_en_2006-06-21.pdf – Tetsujin Jan 10 '22 at 15:10
  • @Tetsujin I did everything as the guide said, in Silverfast's setting, I set input => my calibrated icc and internal => ProPhoto, and the embedded ICC profile checked. When I scanned the image, it produced a tiff file with an embedded ICC profile. So I think the calibration icc was included in the embedded icc? The png files that I uploaded were exported from photoshop for smaller resolution. On the export option, I have convert to sRGB unchecked and embed color profile checked. – reddy Jan 10 '22 at 22:44
  • You're losing the profiles somewhere, perhaps at import. Wherever it is, it's negating your previous efforts to colour manage. As neither of your posted file sets have had a scanner profile, there's not a lot else I can say. – Tetsujin Jan 11 '22 at 09:16
  • @Tetsujin What does the scanner profile look like? Do you mean that when I scanned my image as tiff, the tiff would have a scanner profile and an embedded icc profile? – reddy Jan 11 '22 at 09:26
  • I've never used one, that's why I sent the instructions link. Whatever device generates the image [camera, phone, scanner…] should embed its own profile. Once that's done, other apps can translate to accurately preserve colour. All I can say is that doesn't appear to be happening, or it's being discarded at some other point in your workflow. – Tetsujin Jan 12 '22 at 18:59
  • @Tetsujin From what I understand from the documentation, the scanner first creates the data with its own rgb space, and a scanner/calibration profile and an internal color space declaration will be applied to the data to transform it into image data (in XYZ/Lab space). Then an icc profile can be optionally embedded to the data for proper rendering of the image. The image data was the product of the scanner profile application. – reddy Jan 13 '22 at 07:47

2 Answers2

2

If you are getting the same readings from the same colors on authentic IT8 calibration charts, then the problem is likely something to do with the "paint we both have" and not the scanners.

Batches of paint are not exactly alike; and was the paint applied to exactly the same color/brightness of substrate; at exactly the same thickness/density?

Steven Kersting
  • 17,087
  • 1
  • 11
  • 33
  • Hi, I have like five cels with the same paint on my end, and their values all matched when I scanned them. The difference in hue may be 1 or 2 degrees, but not 10. – reddy Jan 09 '22 at 17:45
  • @reddy "one paint color that we both have" sounds like you're each scanning a sample of "red", but that they're different samples. Maybe they're not as identical as you think. What happens if you each scan the same sample? – FreeMan Oct 07 '22 at 18:16
0

First of all, a hue difference alone is not meaningful because it's an angle revolving around neutral grey (achromatic). For a color of low saturation (near grey), you can have a large hue offset that will result in a very small delta E at the end, because both records will be very close from achromatic (so, large angle with small radius). The only meaningful way to compute the difference between 2 colors is using the delta E.

Then, from your scans, I can see that the target #2 has a warmer grey, so I guess the first issue is with white balance. The expected result of this is more saturated reds and oranges, less saturated blues. But the target #2 is more saturated everywhere, so there is more likely another problem on top.

The fact that both scanners may not have the same gamut changes nothing to the output. In theory, it could, but both your targets are printed on Fuji Crystal Archive Supreme, which is a photochemical paper that will not be able to print colors far outside of Adobe RGB. Whatever gets printed on this paper will most likely fall inside both scanner's gamut.

I would probably check into the profiling software used, and how it is configured.

Aurélien Pierre
  • 827
  • 5
  • 11