How to scan print film 

or: I'm right and you're wrong and nobody cares 

September 8th, 2015
  1. Set your scanner into a known, repeatable state for scanning transparent materials.
  2. Scan a color target using these settings.
  3. Create a profile which characterizes your scanner. I like to use the open-source Argyll CMS software. Matrix or LUT, it's your call.
  4. Scan your negatives using the same settings you used when you scanned your target.
  5. Open your image in your favorite image editor. It has to let you do things like assign and convert color profiles.
  6. Assign the profile you created in step 3 to the image. Maybe you want to save your image now, with this profile, so you don't get confused later on which profile goes with which image (you re-profile your input devices every now and then, right?).
  7. Convert the profile into a working space of your choice. I like to use ProPhotoRGB because there are some wacky colors in print film that we need to preserve accurately. Therefore, when doing this conversion, please use Absolute Colorimetric rendering intent and do not use black point compensation. Other settings skew in-gamut colors in some way, and that is most certainly not what we want at this stage.
  8. Now, finally, you can invert your image. It's kinda cyan-blue! (Your colors will become an insoluble mess if you try to invert your image before converting from your scanner's color space into an abstract working space.)
  9. Create a levels layer (if you can make layers at all, otherwise I guess you're gonna be freeballin' from here on out) and use your favorite techniques for setting a white point and a black point, and finding a neutral mid-tone.
  10. Etc. Everything from here on out is icing.

If you're using VueScan, sorry. You cannot use the color negative setting because the software uses a well-intentioned trick for scanners that support individual exposures for each color channel: it bumps up the exposure for blue and green in order to maximize bit depth usage for those channels, and places the scanned image's colors closer to a realistic palette. It's a nice idea, but the gains are minimal, it can vastly increase scan times, and it doesn't allow you to profile your scanner.

This brings us to the crux of this post and my scanning technique. You will read zillions of confused and confusing posts and articles which often confuse the issue of scanner profiling and film profiling, many of which conclude by stating that one shouldn't concern oneself with color profiles at all when dealing with print film. They then provide workflows which place "raw" scans into an arbitrary color space, such as sRGB or Adobe98 or whatever. This is absurd and makes the process of extracting natural, believable colors from your print film pretty much impossible.

Although my rationale for this technique is unproven by math-y color science (believe me, I wish I had the skills, but you guys: color science is complicated), years of experience of wasted time trying to find a repeatable process that leads to consistent and pleasing results leads me to believe that I'm at least on the right path. The basic idea that informs every step of my technique is that there is a meaningful relationship between the wacky orange/green colors on your print film and that it is therefore important to accurately preserve these relationships up until inversion and color correction. It's just as important as it is for diapositives, and probably even more so because of the magnitude of contrast expansion that has to take place in order to "decode" the colors.

The point of characterizing (profiling) your scanner is so that the color in scanned images has some connection to objective, real-world colors. That source of objectivity is the color target in conjunction with the data file that accompanies it, which contains information about the measured color of every patch. The numbers your scanner spit out have no meaningful relation to any color space or device without a correctly constructed profile. A superficial examination of the RGB response curves in the profile of my scanner indicates that it has some quirks that should not be ignored: non-linearities that result in strange color shifts that are impossible to correct using the usual tools in Photoshop.

Assuming we've compensated for any scanner irregularities as best we can, through the correct application of a scanner profile, it is important that we do not alter colors in any way when moving from the scanner color space to our working space (as we might very well do when when moving between working spaces, for example, or to an output color space like a printer). I wish to apply a rendering intent that does not shift the white or black point (Absolute Colorimetric without black point compensation); and if possible, I wish to move into a work space that can fully contain the gamut of the negative film. The only readily available working space I have experimented with that fully contains the gamut of my negatives is ProPhotoRGB.

Believe me, I'd love to work in a smaller color space at 8 bits a channel. Everything would be faster and I'd need to use less disk space. But the fact of my observations is that compressing or clipping the color of my negatives into makes it incredibly difficult to effectively remove color casts across the entire tonal range of my images, because any shifts introduced by color space conversions, however well-intentioned, are immensely and seemingly unpredictably magnified when you set about expanding and color correcting your negatives.

By preserving the color relations on the film by assiduously observing this workflow, I'm able to avoid most mystery color casts. I have found that it is possible do the great majority of color corrections to my scanned negatives using only the Levels tool in Photoshop. If you find that you're having to use really lumpy curves, and different curves in each color channel, in order to get only mildly acceptable results, I would strongly suggest that you try my technique.