Color Correction Advice

Discussion in 'Digital Darkroom' started by dfperrault, Jul 17, 2021.

  1. In which case that grey scale and those swatches are useless.

    Removing the paper yellowing reveals the colours more as they would have been originally, but without knowing the hue and intensity of the original inks, a faithful restoration isn't possible.

    Here's a quick edit that restores the base paper (mostly) to near-white.
    IMG_20210808_101534.jpg
    You can see that it makes a nonsense of any external 'reference' colour.

    So the choice is eye-and-emotion versus instrumentation. Your choice, but personally I'd go with the Mk1 eyeball.
     
  2. digitaldog

    digitaldog Andrew Rodney

    Indeed, as outlined in post #8 (Color Correction Advice) ;).
     

  3. Is this data in this article sufficient or would your measured values be more accurate. What are you rates?

    https://www.imaging.org/site/PDFS/Papers/2002/PICS-0-267/7122.pdf

    Do I fine tune the need to manually create a calibration file or try to fine tune the RBG slider to match the image target vales to the reference values. or do I need to do a Curve fit and change gamma settings within photoshop?
     
  4. Rodeo Joe and others are right in questioning whether you need the kind of precision you seem to be asking for, for your purposes.

    His Q&D tonal adjustment can get you close enough for a starting point, and then you can play with adding 'filters' in a reasonably complex image manipulation program to get something that "looks' right to that mark 1 eyeball
     
    digitaldog likes this.
  5. From that linked 'scholarly' PDF article: -
    "ChannelValue = 255(Reflectance 1/γ )"
    Except that the sRGB colour space doesn't actually use a true gamma 2.2 tone curve, and nor do any other standard spaces use a true gamma of 1.8 - close, but not scientifically accurate.

    Only AdobeRGB uses a true 2.2 gamma curve, but I suspect sRGB was what the authors of that old article had in mind with their table of gamma 2.2 values.

    For example: A reflectance of 0.011 should equate to pixel value 27, 27, 27 in sRGB space. Not 33 as given in that table. The reflectance value 0.178 (patch 7-M) also has its closest pixel value at level 117, not 116.

    OK, for the majority of greyscale patches the pixel values aren't that far in error - theoretically. But the theory takes no account of the inevitable flare and degradation of reflective values actually captured by the camera or scanner reproduction. And therein lies the rub!

    IMO, any attempt to apply the 'correct' tonal or colour values given in that highly theoretical paper are doomed to failure under real-world conditions.
     
  6. digitaldog

    digitaldog Andrew Rodney

    The original sRGB has never used a gamma curve. It uses a TRC; they are not the same. A TRC may follow the gamma formula or may not. A gamma curve always does.
    http://digitaldog.net/files/LinearityandGamma.pdf
    Also, Adobe RGB (1998) among other Working Spaces use a true 2.2 Gamma Curve. It isn't unique in that respect. So only? No sir.
    Oh, when discussing sRGB, it sometimes helps to define which sRGB:
    sRGB profile comparison
    See Table A2.
     
  7. They calibration target is in every image so I thought I would try to use it. If there were all photographed at the same time / conditions, I can apply the correction for one target to all. so it is not that difficult. This would eliminate the ambient effects so I can focus on color matching to high quality text books Images.

    Also what is "Q&D"
     
    digitaldog likes this.
  8. If I want to mathematically match the colors/exposure of a target image (image to be corrected) to a reference image, could I do this by calculating the difference or ratio of the two images in PS, and scale the target image to match the reference image. I would have to make sure they are physically scaled to the same size /alignment. Would it be better to do each color channel independently in RGB of CMYK?

    Method 1 - Differential Error
    Layer 1 - Reference Image
    Layer 2 - Target Image
    Layer 3 = (Target - Reference) / Reference
    Layer 4 = Target / (layer 3 +1)

    Method 2 - Ratio
    Layer 1 - Reference Image
    Layer 2 - Target Image
    Layer 3 = Target / Reference
    Layer 4 = Layer 2 / Layer 3


    If this method works, then I could do this in Matlab, and have all the 100 images processed in a minute or so. But I suspect it is not that easy.
    BTW, I just realized what Q&D means
     
    Last edited: Aug 29, 2021
  9. digitaldog

    digitaldog Andrew Rodney

    Don't take this the wrong way but...
    There seems to be some perverse human characteristic that likes to make easy things difficult.” -Warren Buffett
    You could however, fix one image visually, doing this all on an Adjustment layer, then drag and drop it on all the others.
    Or create a preset of the fix visually and copy/paste them to all others in say Lightroom/ACR.
     
  10. Sorry, I have not explained my goals in enough deals. I started off posting a color correction question based on a target.

    There are two levels of adjustment I need to do
    1. Color correction for ambient conditions (this is not that difficult)
    2. Color/Exposure restoration of the target print to match a published museum quality version

    Each of the 100 target images have been collected from different sources, storage conditions, different print dates etc. over the past 150 years. The target Images need to be individually corrected to match the higher quality version.

    Sure, I can do this by eye for the 100 prints. 25-50hrs of work. Just curious of there are some computational solutions that are less subjective.
     
  11. "Q&D" - Qick 'n' Dirty. Which I admit my quick restoration of the paper base to near-white was.

    Unless you have access to 16 bit, or at least 12 bit images, then doing the adjustments in two stages isn't a good idea. And what good does it do you to get the base paper to its currently 'correct' shade of yellow or dirty brown? When you'll only have to reduce the red/green channels, or boost the blue, or both, to get close to the unfaded and un-aged look the artist created.

    If you want automation: Try cropping the images to exclude everything except the drawings and simply hit 'Autocolor' in Photoshop.
     

Share This Page

1111