You have a number of things going on here. The biggest problem is the variety of colorants in the original image. Or to be more specific, the different spectral power distribution of the colorants. Two "colors" made of different physical colorants may appear to match under one illuminant but not under another. Or they may appear to match to one detector system (e.g., the human VS) and not to another (e.g.m film, scanner, digital camera). These are all very simple tristimulus detection systems that cannot take the necessary multitude of spectral "slices" of the sampled colors. UV fluorescence is a related problem--very common in photographing (and scanning) artwork. Filter the light source if possible. Paper color in older documents may not be a difficult issue as they tend to lack both strong color or UV fluorescing agents (optical brighteners). Just don't "white out" on the substrate. If you print render with absolute colorimetric intent, like a proof. I don't see any fail-safe method for accounting, through ICC profiling, for every combination of pigments and dyes you'll encounter. Can you make a profiling target for each of these, using the actual (obscure, ancient) colorants in the profiling target? Of course not. Now if you had a spectral multisampling capture device instead of the usual RGB systems the translation could be done. Maybe Nasa has something... I haven't looked closely at the system that HP marketed recently--my recollection is that it did make use of some limited spectral sampling of the artwork. Does anyone know more about this? I seem to recall it was tied to a Z-series printer.
Message: 1 Date: Thu, 21 Jul 2011 17:28:46 +1000 From: "mail@iccimagetec.com.au" <mail@iccimagetec.com.au> Subject: Archival image reproduction on Kodak Creo IQ3 scanner To: ColorSync User List <colorsync-users@lists.apple.com> Message-ID: <E7D5E88E-0853-4BD1-9B95-B15C73580865@iccimagetec.com.au> Content-Type: text/plain; charset=us-ascii
Hi I am sure there are many on this list who have had to deal with reproduction of historical documents for state or national archives. I have a client who is using a Kodak Creo iq3 scanner with OxyGen software to scan documents ranging from 19th century glass plates, albumen and silver gelatine photographs, water colour and ink illustrations, hand coloured B&W images right up to modern day emulsions. We have built two profiles in ProfileMaker. One based on a reflective photographic target for standard emulsions and one based on a Colourchecker SG target for other reflective media. With a small amount of tweaking in Profile Editor we can get the white, grey and black patches on a X-Rite greyscale checker to match the target RGB values in Adobe RGB within one level.
My question is, is there a subject based spectral response to the scanner light source that will render each of these types of original material differently with respect to the colour checker patches. In other words, are we likely to have to build a material specific correction for different types of originals? We have found, purely visually on a small range of samples, that there seems to be one type of cast for one material and a different cast for something else.
Without going into specifics, is this an accepted issue and would a workable solution be scan all reflective originals with the same scanner profile and run a different photoshop action as a correction layer for each different type of material. This would seem simpler than having 10 different edited versions of the scanner profile, and would be easy to adjust over time as more samples are taken into account.
Thanks in advance.
Simon Olding ICC Imagetec mail@iccimagetec.com.au www.iccimagetec.com.au photo galleries: www.simonolding.com P: (03) 6223 7882 M: 0424 022 831