| Re: Image correction for human perception? [message #50428] |
Mon, 02 October 2006 10:57 |
b_gom
Messages: 105
Registered: April 2003
|
Senior Member |
|
|
Hi Maye,
> But after we have our calibrated images in the 3 colours and put this
> together to a colour image, why would we have to correct for the human
> eye spectral sensitivity, if looking at the image at the screen
> (provided my display system works ideal) will automatically involve the
> eye's sensitivity?
When it comes to color science, 90% of the problem is terminology.
OK, assume you have a perfectly calibrated spectrometer that gives you
the true spectrum of the image across the visible band, or you have a
good 3-color CCD system with known filter characteristics. Now, the
goal is to put something on a screen that looks like the original scene
to the average observer.
In principle, all you need to do is figure out what the relative
brightness each of the 3 cones in the eye would register when looking
at your perfect spectrum. CIE has established some reasonably accurate
spectral sensitivity curves (color matching functions) that allow you
to do just this. Now you just need to take these 3 values and pass them
through your perfectly calibrated video card and ideal display. In
reality, this means you have to correct the image for the limitations
of the software and hardware in order to produce an output that is
reasonably close to what the original spectrum would 'look like'. This
is where things like 'rendering intent' come in and complicate the
whole color correction process.
Of course, there are other problems too, such as how room lighting
actually affects what colors you think you see. For example a sheet of
paper looks white under incandescent light as well as in sunlight, even
though it is actually more 'red' in the first case. This is true for
monitors as well.. Also, not everyone is the same as the CIE standard
observer.
As an exercise, take a digital picture out your window. Load it onto
your monitor. If you're really lucky, it might almost look the same
color as the original. Now put a second monitor next to the first, and
display the image on both. Your luck has probably run out. Which one is
the true color?
Check out http://www.normankoren.com/color_management.html for a good
summary and links to other color science documents.
Of course, all of this is completely different from the problem of
displaying data for either colors or intensities that the eye cannot
perceive..
Brad
maye wrote:
> Hi!
> We have an interesting discussion going on here about the need of
> correcting intensities of remote sensing data in red, green and blue
> filters for the perception of the human eye to create a 'TRUE' color
> image.
> Basically, at the moment we tend to believe it's not necessary, because
> we have a good linear detection system (a CCD),
> so all we need to correct is the filter absorption (plus the usual CCD
> tralala of course).
> But after we have our calibrated images in the 3 colours and put this
> together to a colour image, why would we have to correct for the human
> eye spectral sensitivity, if looking at the image at the screen
> (provided my display system works ideal) will automatically involve the
> eye's sensitivity?
> Would the best way to display remote sensing data to the human eye not
> be, to try to show the exact same relative intensities like detected at
> the place of observation?
> What makes us doubt is the amount of publications one can find
> mentioning a "correction for human eye perception" and we fear, that
> they cannot ALL be wrong! :)
> So what do we miss? Is the problem maybe, that we only have data of 3
> filters and there's of course more? But the television works like that
> and can create quite realistic images, so it shouldn't be the problem?
> Thanks for your enlightenment!
> Regards,
> Michael
|
|
|
|
| Re: Image correction for human perception? [message #50447 is a reply to message #50428] |
Mon, 02 October 2006 08:15  |
edward.s.meinel@aero.
Messages: 52
Registered: February 2005
|
Member |
|
|
maye wrote:
> Hi!
> We have an interesting discussion going on here about the need of
> correcting intensities of remote sensing data in red, green and blue
> filters for the perception of the human eye to create a 'TRUE' color
> image.
> ...
> What makes us doubt is the amount of publications one can find
> mentioning a "correction for human eye perception" and we fear, that
> they cannot ALL be wrong! :)
Sure, they can all be wrong (it wouldn't be the first time...). I
really hope that those were not scientific journals, because the phrase
"correction for human eye perception" is meaningless without specifying
the "correction." My guess is that the dynamic range of the image is
too great to display properly in the publication so they tweak the
image to make it more visually pleasing.
As for displaying a TRUE true-color image, that can only happen if the
spectral response of the detector AND display match that of the eye.
Normally, you won't even get the display to match the spectral response
of the detector, which can lead to some rather odd visual effects.
Ed
|
|
|
|
| Re: Image correction for human perception? [message #50454 is a reply to message #50447] |
Sun, 01 October 2006 05:03 |
George N. White III
Messages: 56
Registered: September 2000
|
Member |
|
|
On Sat, 30 Sep 2006, maye wrote:
> Hi!
> We have an interesting discussion going on here about the need of
> correcting intensities of remote sensing data in red, green and blue
> filters for the perception of the human eye to create a 'TRUE' color
> image.
> Basically, at the moment we tend to believe it's not necessary, because
> we have a good linear detection system (a CCD),
> so all we need to correct is the filter absorption (plus the usual CCD
> tralala of course).
> But after we have our calibrated images in the 3 colours and put this
> together to a colour image, why would we have to correct for the human
> eye spectral sensitivity, if looking at the image at the screen
> (provided my display system works ideal) will automatically involve the
> eye's sensitivity?
There is more to the process of viewing images -- you have to consider the
limitations of the output device. Think about conventional photography
where skilled photographers vary the exposure time for different areas of
a print to bring out shadow detail that would otherwise be lost.
I work with "ocean colour" where we are interested in the subtle
differences in the light leaving the water surface resulting from
differences in the chlorophyll concentration (which can vary from 0.01 to
100 mg chl m^-3) among other factors. If you consider the earth as a
whole, with cloud and land, the oceans are dark and you don't see many
features. Yet it is possible to produce dramatic false-color images of
chl concentration (see http://oceancolor.gsfc.nasa.gov/)
> Would the best way to display remote sensing data to the human eye not
> be, to try to show the exact same relative intensities like detected at
> the place of observation?
> What makes us doubt is the amount of publications one can find
> mentioning a "correction for human eye perception" and we fear, that
> they cannot ALL be wrong! :)
> So what do we miss? Is the problem maybe, that we only have data of 3
> filters and there's of course more? But the television works like that
> and can create quite realistic images, so it shouldn't be the problem?
> Thanks for your enlightenment!
Some applications correct a scene to what you would get with some
"standard" illumination (e.g., local noon on a cloudy day), especially
where multiple images are "stitched" together.
Start with http://www.cis.rit.edu/mcsl/iCAM/ and http://debevec.org/
--
George N. White III <aa056@chebucto.ns.ca>
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
