Sensor filters
Sensor filters
- Subject: Sensor filters
- From: Robin Myers <email@hidden>
- Date: Fri, 26 Jul 2002 14:42:26 -0700
- Organization: Robin Myers Imaging
>
This is my first posting to the list, being an imaging science student
>
preparing a dissertation on digital reproduction of paintings. I understand
>
that the ccd's in scan backs are often based on scanner technology, and as
>
such are optimised for cmy (dyes) color accuracy, not rgb. This would mean
>
good color accuracy for photographic artwork, but not for pigments. An area
>
array ccd in a digital camera will give far better color accuracy. Imaging
>
scientists describe the two different responses as 'densitometric' (cmy) and
>
'colorimetric' (rgb).
There is much misinformation about scanners and digital cameras. The
selection of the peak responses for the filters was not made to optimize
their response to CMY inks. The reason is much simpler.
The sensor filters traditionally were selected by dividing the visible
spectrum of 400 to 700 nm into three regions corresponding to 400-500,
500-600 and 600-700 nm. The filters were then designed to have peak
sensitivities at the center of each of these regions; 450, 550 and 650
nm. By doing this, the conversion matrices for RGB to XYZ are often
simplified to 3x3 matrices, making conversion fast and simple. It also
simplifies some operations for color constancy calculations.
Scanner filters were not designed to be optimized to CMY film dyes or
printing inks. In fact, they need to respond to a very wide variety of
colorants since they need to scan almost any type of material placed in
the scanner (some people I know place even place fabrics into their drum
scanner for imaging). Densitometer filters are tuned to the spectral
response of printing inks and film dyes, but these spectral responses
are different than those of the filters used in general purpose scanners
and digital cameras. They were simply designed to make it easy for the
engineers to 'do the math'.
>
Some pigments are known problem colors for (particularly digital)
>
reproduction. I would be very interested in hearing about list members
>
experiences of problems with getting accurate color.
Interestingly, the 450 and 550 peaks are very close to the human visual
response peaks. The 650 peak response is the one that causes the most
problems. The human long wavelength response peaks about 600 nm, a
red-orange wavelength. Many dyes and pigments have high reflectances at
650 nm and longer wavelengths; where the human visual system has low
sensitivity and the digital sensor has high sensitivity. The paper
"Color Accurate Digital Photography of Artworks" in the Information area
of www.rmimaging.com explains in detail this observer metamerism problem
and its solution for digital cameras.
Be careful not to paint pigments (excuse the pun) as the main culprit in
color reproduction problems. The list of dyes that cause color
reproduction problems is a far longer one than the pigments. There are
certain dyed fabrics that are well known amongst commercial
photographers that are extremely problematic (certain greens, reds,
blacks and blues, especially). I have databases with thousands of
spectra of flowers, inks, paints, plastics, papers, dyes, etc. and most
of the reproduction problems are with the long wavelength response (600
vs. 650 nm).
As for the difference in the spectral responses of filters used in
digital cameras and digital scanbacks, there is very little. Look
carefully at the specification sheets for trilinear (scanback, scanner)
sensor vendors such as Kodak and Sony, then look at area array (digital
camera) sensor responses from vendors such as Kodak, Sony, Phillips,
etc. and they have almost the same peak sensitivities for the same
reasons listed above.
Robin Myers
_______________________________________________
colorsync-users mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/colorsync-users
Do not post admin requests to the list. They will be ignored.