Re: Color of a Rose
Re: Color of a Rose
- Subject: Re: Color of a Rose
- From: Robin Myers <email@hidden>
- Date: Thu, 4 Jun 2009 11:06:37 -0700
Hello Don,
The Foveon sensor is a tri-layer sensor. Instead of pattern
interpolation like the Bayer pattern sensors, they must do
deconvolution to remove cross-talk between layers. Their technology
works based on the wavelength dependent penetration of photons into
silicon. Longer wavelengths penetrate deeper. So the layers are blue,
green then red in order from top to bottom. The last time I looked
there was a filter layer in there somewhere to improve the separation
between layers.
Robin Myers
On Jun 4, 2009, at 10:38 AM, Don Hutcheson wrote:
1. The red rose is probably outside the color gamut of either the
camera, the monitor, the working space, the imperfect 8 bit
interpretation of CIElab, or a combination of all these. If so,
normal ICC color management isn't going to solve the problem.
2. If the rose is outside the camera's sensitivity gamut, in the
original image (exported from ACR into ProPhotRGB) some pixels (in
the red rose area) should have zero values in the green and/or blue
channels and/or 255 values in the red channel. In either case a
zero or a 255 indicates probable clipping. If there are no zeros or
255s, the camera is POSSIBLY not at fault.
3. If you can edit the rose manually in Photoshop (by whatever
means) to look convincingly like the original rose, or at least a
lot closer, then the monitor may not be the whole problem.
4. If both above tests pass (i.e. no apparent clipping in camera and
ability to get a better match manually on the monitor) then the
culprit may be metamerism failure between the rose red and the ACR
camera profile in that color region. Such errors are to be expected
in saturated colors and might be due to invisible wavelengths (IR or
UV) seen by the chip but not by the eye, or a host of other causes.
5. Although plants may not fluoresce they certainly do reflect UV
and IR light. Insects can distinguish between flowers that we see
as the same color due to UV and IR reflectance levels. Near-IR is a
known cause of metamerism problems in fabrics, even though a typical
lens glass is a pretty good blocker of far-IR wavelengths. Adding
near-IR or near-UV filters at the time of capture might help.
6. Somewhat off-topic but raised by this thread; Foveon's marketing
blurb claims that their sensor is multi-layered, like a tri-pak film
emulsion. However Foveon's patent diagrams appear to contradict
these claims, showing all three RGB sensors on the surface of the
chip, presumably in a pattern of concentric circles or rectangles.
In the cross-section patent diagram the blue sensor appears to be a
central "point", surrounded by a green sensor ring, in turn
surrounded by a larger red sensor ring. The red and blue rings are
probably more rectangular than circular for efficiency sake.
If Foveon's sensor is in fact planar (single layer) as the patent
diagram implies, and not tri-pak as implied by Foveon's marketing,
then it is still pattern-based and thus not immune to possible color
artifacts. Although Foveon wouldn't have to apply a traditional
Bayer de-mosaicing algorithm, the three sensors would not in fact be
seeing exactly the same image. If the subject included very fine
high-contrast granular detail, colored artifacts might result - for
example a small black dot covering the central two sensors but not
the outer ring would produce a red pixel, not a black one.
Anyway, if Foveon's sensor is, as I suspect, just a modified a 2-D
pattern on the surface of the chip, then in fact a Foveon chip
separates color via a layer of RGB filters printed on the surface,
just like a Bayer-pattern sensor, and there should be no inherent
difference between how a Foveon-patterned and a bayer-patterned
sensor "see" color, except for the choice of RGB filter colors and
any 3x3 matrix applied before the "raw" RGB signals are exported.
Bottom line: This problem is most likely due to camera clipping or
metamerism failure due to IR or UV reflectance. If so you will have
to edit the image manually.
Don
........................................................
Don Hutcheson
HutchColor, LLC
Washington, NJ 07882, USA
office: (908) 689 7403
cell: (908) 500 0341
........................................................
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden