Hi Stefan,
I'll try to answer your <<questions>>, based on my experience
with this approach, but perhaps its becoming a bit specialized for general
interest and we should take it off line hereafter.
<<If I read you right, it would be possible to just use 3
monochromatic
sources of light with a known
wavelenght of red,green blue
to readout all the spectral response
informations of this (for example
Bayerpattern) chip.
The mixture of colors as well as the simulation of
"greys" would need
specific modifications of the emissive sources
which
may in result not improve the readout of specific values, but in the
opposite diminish the exactness of the measurements ?
Because as we have
discussed here recently even high end measurement devices
that would be
needed to determine the exactness
of the mixtures will have a certain degree
of misreading. Adding up these
errors (system noise!) will most likely
result in a loss of profile quality >>
No. You really need to run a complete scan across the wavelengths of
sensitivity of the camera. Essentially, you want to measure the color matching
functions of the camera - instead of x-bar, y-bar, z-bar, you measure
r-bar, g-bar, b-bar for the camera. With this information, the linearity
characteristics, and a few other basic characteristics, you know how the camera
"sees" and can quantify the camera colorimetry. This actually leads to more
accurate profiles with lower noise propagation.
<<But wouldn´t this target be extremely expensive and also hard to
handle ?
All the fileds values have to be determined spectrally in a Lab
environment.
Any flare or reflections on the surface
of the target will
completely destroy the reliability of this approach.
What happens if the
environments brightness exceeds the brightness of the
targets emitted
lumen?
So it probably could only be used in the dark....?
As the Luminance
of the target will be produced by LEDs (?) with electric
power sources
,
how stable is the LED spectral output if voltage goes down ?>>
Yes, a monochromator is very expensive and so is the person to run it.
Which is the potential beauty of the LED target. Of course target
presentation is important for camera profiling. But the LED target is probably
easier to present properly to the camera than a reflective target It would
be shot in dim or dark conditions eliminating the ambient problems common with
reflective targets. LED-based systems can indeed be extremely stable with
simple, inexpensive feedback and control circuits. The LED target could be
supplied with its own characterization data as is commonly done for other
targets, but I'd guess its unnecessary for most applications.
<<As you mention Lens Flare, would this not add wavelenght specific
faults
according to the
lens specifics, instead of specifiying single
color measurements of the chip
?
I mean there will be totally different
values for 12 Lens Zooms or Macro
triplets-
further should a Zoom be
measured with the shortest or the longest focal
lenght
?........>>
As I mentioned before: In principle, yes, similar to other profiling
methods. But, in my experience, if you really have good camera characterization
data and a well-optimized camera profile, it can be surprisingly robust in the
face of small changes in system setup. Of course, if the spectral and linearity
transmission of the photogs lenses vary greatly, the LED target is a potentially
excellent way to profile each permutation in his setup.
<<Now what about chip heating, aging, environment temperature and
software changes? What does the influence of ASA settings/speed as well as
exposure times mean ? Does this mean a photographer should continuously
recalibrate with the emissive target ?>>
Of course these are the same issues faced with reflective targets. You are
profiling the entire chain. But with the mono-based approach, the situation is
ameliorated somewhat more robustly. With the reflective target, the profile is
highly optimized to the training data which is far noisier from a system
perspective.
<<... a target means nothing but additional trouble
producing
intense handling needs with very limited use for outputting real
Photographs. As I see it, this is mostly a scientific Lab solution, normal
photograhic users will have BIG difficulties to achieve any usefulness of
this.>>
Unlikely. I have just the opposite experience with these kinds of devices.
Far easier handling, far more robust results, and quality approaching laboratory
results at a fraction of the cost. Of course, it remains to be seen how well the
commercial implementation is performed.
I have no skin in this game, But if the first commercial release is
half as good as its potential, I, for one, will be a happy customer and
enthusiastic evangelist.