stevebow wrote:Hi Arne,
Could you please expand on this? I have often wondered why IT8 profiling is not considered when scanning negs.
Let's say my aging scanner is loosing sensitivity in the red channel. This is very obvious visually when scanning a positive image, and an IT8 calibration will compensate for this problem.
A scanned image is a scanned image, regardless of the image source type (neg/transparency), so the "colour" of the orange mask of a given scanned negative will look different scanned when the scanner is old compared to new. If SF Ai does not use IT8 calibration when scanning negs, how does it compensate for image sensor variation/degradation over the scanner's lifetime?
Steve
I think the reason is that with a positive image the image itself is the absolute reference of how it
should look and so a scanner just has to faithfully replicate what is there. The scanner is already designed for this purpose and a custom IT8 profile attempts to correct for any deviations by measuring a handful of know colour patches to modify the scanner's response. (I think this is flawed too and will present my case below!)
A negative is obviously not the final image since it comprises clouds of cyan, magenta and yellow dyes behind an orange mask. The incorrect assumption often made is that if you can faithfully replicate the cyan, magenta and yellow image (combined with the orange mask) then you can faithfully replicate its inverse. There is no industry standard method of removing the mask and inverting a negative. This was originally done optically during the wet printing process by a technician skilled in adjusting the CMY filter packs to suit. Every film stock is different and may vary further according to exposure, development and storage methods.
If you open the Advanced part of NegaFix on one of your scans and click the Expansion tab you can see the fairly narrow space occupied by the 'data' on the film. Firstly you can see how low contrast colour negatives are, which is added later according to the photographic paper used (optical printing) or in software (scanning). Secondly you will see that to remove the mask the delimiters at the start and end of each histogram are numerically all at very different places, as are the histograms themselves in relation to one another.
It doesn't really matter if any given colour channel loses some sensitivity because you (or the software) still has to go through the process of determining the start and end points of each histogram and their final positions relative to one another, which negates any issues of where the histogram starts out - there is plenty of empty space either side!
With regard to scanning slides, I have had infinitely better results turning colour management off. Initially I get a red cast in shadows, suggesting the scanner's red channel is fairly dominant but one click of Levels in Photoshop restores the correct relationship and the colours are far superior to an IT8 colour managed scan. Why? The IT8 target has a handful of known colour patches from which it constructs an RGB matrix whereby any other colour can theoretically be interpolated. But the colour space is now warped instead of being highly linear, as the scanner is without colour management. While the RGB primaries of my scanner probably don't match that of the Adobe RGB colour space that I
assign to the scans, the colours nonetheless retain an accurate
relationship between each other, and all I can say is they look much, much better. I'd be happy to link anybody to my Flickr page via PM so you can see what a home scanner can do without going through the pain of IT8 profiling.