On 08/12/2013 09:48 AM, Michael Darling wrote:
Out of curiosity, I measured my scanner's bulb's light spectrum (spotread -a -H), at night, all other sources of light turned off. It was quite interesting. (Closest standard illuminant would be F10, but there's significant differences.) Is there anything useful that I can do with this spectral data to improving the color accuracy of my scans? I looked at the colprof, spotread, and .SP file documentation, mostly having to do with the -i switch. I'm thinking this only has to do with illumination on the output side, not the input side... But want to make sure I'm not missing something! I'm thinking the profiling process effectively makes the bulb's spectral data irrelevant... By comparing the values given by the scanner to the spectral data of the target, it basically takes the bulb's spectrum out of the equation, right? (Obviously without removing the limitations of the bulb's spectrum like there's an extremely small amount of 520nm light, so there's going to be trouble there.)
Set that spectral plot of the scanner light against a plot of the scanner sensor filters. And a plot of the CMY dyes of several emulsions. Many scanners, even desktop models still aim at (or are limited to) reproducing photo originals. Within the Epson desktop range the V700-V750 target photo originals (including film) more than the Epson 3200 did. My experience. You have to know the sensor filter dyes too to get a picture of the scanner's limitations. I doubt that knowledge could contribute to a better CM solution, the limitations in total filtering remain. Conventional printing CMYK pigments and Officeprinter CMYK dyes or pigments are close enough to chromogenic CMY dyes. It is harder to get it correct for more pigments beyond CMYK. Few that can scan all the basic pigments of a watercolor paint box correctly, not to mention real life colors. The continuous spectrum of a Xenon light source should improve on that but few scanners have that these days.
In the HP G4010 and G4050 two lamps are available with different spectral distributions. In some driver modes two scans are made, each with a different lamp. The two are combined where one delivers the detail + color and the second one only compensates the color of the first scan by some smart algorithms. My guess is that it even can create more color fidelity than if it made one scan with both lamps on. Based on the different metameric effects created per lamp.
http://nl.hardware.info/reviews/1044/2/hp-scanjet-g4050-review-6-kleuren-technologie There could be an English version. http://www.image-engineering.de/library-m/test-reports Third PDF on that page. -- Met vriendelijke groet, Ernst Dinkla http://www.pigment-print.com/spectralplots/spectrumviz_1.htm December 2012: 500+ inkjet media paper white spectral plots.