Re: Primary colors


"Chris L Peterson" <clp@xxxxxxxxxxxxxxxxxx> wrote in message
news:0vemd2dkc8d3o0eae1ttus3lcp4s0r2771@xxxxxxxxxx
On Thu, 10 Aug 2006 10:43:00 GMT, pausch@xxxxxxx (Paul Schlyter) wrote:

To bring things back to astronomy, one of the most complex color theory
problems involves converting between spaces. Some of our space probes
have high order color spaces- images may consist of intensity data in 10
or more wavelength bands, some of which may be narrowband (not
overlapping adjacent bands). It is a challenge to take such data and map
it to a trichromatic space like RGB. Perhaps you recall how long it was
before Mars Rover images finally started being released showing
approximately "true" colors, and even now many are never converted.

Yes, astronomy and color are challenging to mix!



One must remember that color is a combination of three things: illuminant,
and object, and an observer. Sometimes just two: something that glows, and
observer. The problem with space probes is that they are not designed to
capture true color. Rather they are designed for other spectroscopic
analysis or often weight/stability, or data bandwidth limitations.



The real issue is that most imagers capture only sparsely sampled spectra,
or abbreviated spectra. If we had a detailed spectrum, we could
reconstitute the measured color quite easily by using the color matching
functions of the human visual system and obtaining the XYZ tristimulus
values. These can be converted into any other trichromatic color space - of
which there have been dozens defined/created.



But think of the Nyquist sampling theorem: if the rate of sampling is not
high enough [in whatever space you are working] then the true waveform can't
be captured due to aliasing. Thus the Mars Pancam images are multi-spectral
images, but only 6 channel, narrow band. One must make assumptions about
the general shape of the spectrum to convert to XYZ so they are only
approximate true color.



If you assume that the approximation is close enough, you will note that we
have only captured the measured color - we still have to consider the color
_appearance_ to the human observer. There are color appearance models that
attempt to take into account the numerous color appearance effects as
observed in the human visual system to predict how the colors would appear
to a human. This requires one to carefully choose parameters for the model
to match the viewing conditions.



This last part of predicting the color appearance is the most difficult part
of the process of creating color astronomical images. It is further
compounded by the gamut mapping that must occur when converting to/from the
theoretical color spaces to device color spaces of the imaging device
[camera] and the rendering device [printer/display]. When you combine the
limitations of the imager, with the limitations of the renderer, factoring
in the color appearance effects, then you can see that it is indeed an
incredibly complex problem.



---------------------------------
Jim Hewitt, Color Scientist
Color and Imaging Team
Advanced Technology Section
Hewlett-Packard Company



.

Relevant Pages

  • Re: Office Pro applications not capable of correct conversion to p
    ... in RGB - including all text and images, ... to and remains RBG. ... converting to .pdf, preserving color will keep the entire document in RBG. ... Microsoft Publisher also seems to support only RBG. ...
    (microsoft.public.office.misc)
  • Re: converting 2007 pptx into 2003 ppt without text changing to images
    ... still everytime I try and save as 2003 it still converts the text to images. ... A note to microsoft - who's idea was it to make converting text to images ... format to put the file in compatibility mode (and disable at least some features ... Steve Rindsberg, PPT MVP ...
    (microsoft.public.powerpoint)
  • Re: Pentium M differences: 1.73Ghz? 1.86? 2.0?
    ... applications faster than a 1.73 GHz cpu your claim to saving minutes ... converting 100 images would have to be proven to me. ...
    (comp.sys.laptops)
  • Re: Problems Converting Bitmaps
    ... I implemented the OctreeQuantizer but my converted images colors are messed ... The pixel value coming back from QuantitizePixelis 1 so it seems like the ... Tracing through the code it looks like the 32 bpp colors only have one byte ... My program is written in C# and I have tried various means of converting ...
    (microsoft.public.dotnet.framework.drawing)