[Tig] Scans from different scanners

Kevin Wheatley kevin.j.wheatley at gmail.com
Thu Nov 20 10:10:27 GMT 2014

So the goal of scanning colour film is to recover the dye densities at
each position of the film, to do that you want to pass light through
the film, some of which will get absorbed by the dyes (and base) but
let through some frequencies, that can then fall upon a set of sensors
that can distinguish the effect of the different dyes. As part of this
your need to optimize a number of aspects such as noise/dynamic range,
sensor spectral sensitivities, light source power distribution, etc.
You also may consider the typical absorption properties of different
film types (neg vs print etc).

Unfortunately there is no perfect choice available, when trying to
make a cost effective scanner. Ideally you could spectrally scan the
film but this would be impractical as you would need too long an
exposure to get a good signal (you can't push too much light power
through the film without damaging it). Instead you try match the peak
sensitivity of the sensor with the product of the film and light
source (including lenses, reflectors, light guides etc.) so that the
exposure time is lowered and then you can scan faster without melting
the film.

So when you calibrate a scanner you should ideally spectrally scan a
variety of each film stock exposed with a set of colour patches, the
slow and hard way, then repeat the scan with the scanner you want to
calibrate. You can then calculate an ideal result your scanner should
get from the spectral and use that to build a calibration for the
actual delta. As this is a reduction in basis (many channels of
spectral data down to 3 RGB values) this is an approximation. This and
other variations such as exact lamp colour, etc. means that different
scanners produce different outputs even after calibration.

Even worse is because of all these compromises your scanner becomes
sensitive to changes such as lamp switches, filter fading etc. this
means you should re-calibrate more often than you would like.

The ACES printing density is intended to give an ideal target for each
vendor of scanners to target to try make similar scans. In essence it
is a revised attempt to standardise what was done in the Cineon system
and is almost identical to how Cinesite calibrated their scanners but
with modified targets. My understanding of how other scanning vendors
calibrated their systems is limited but I am led to believe they also
followed similar approaches - but may have targeted different sets of
density e.g. Status M, rather than printing density (note this is not
Status A).


More information about the Tig mailing list