posted on Jan, 19 2004 @ 09:21 PM
Originally posted by tuc
I enjoyed the article at www.atsnn.com... but two things occur to me.
1) It would be nice if NASA could tell the parameters of the "hard limiting" algorithm for each exposure. If they could give us the actual low and
high values (in any convenient units) they map to 0x00 and 0xFF, it would allow us to composite them in ways we can't now.
Does NASA even have this data? If not, I guess it would be unreasonably to hope they would share it with us.
AFAICT, you can't truly know these high and low luminance values because they probably depend on the exposure, and MER sends it back already
modified. I don't see that as a bug, though. You have to realize that some colours, like bright sunlight, can't ever be reproduced on your monitor,
much less on paper. The best we can do is approximate a balanced image within the limits of the RGB medium.
What Kano refers to as "hard limiting" is called equalizing in the image processing world. Or at least that's what Adobe calls it. You can see for
yourself what this does to an image by selecting Image/Adjustments/Levels... and then clicking on "Auto." You can see how the histogram levels
stretch to fit the 0x00 to 0xFF scale. The Auto Levels function in the Adjust menu does the same thing, just one less step.
2) It seems to me that it wouldn't be too hard to take the RGB data from an image captured with 750/530/480 filters and translate it to one
approximating what it would look like if it had been captured with 600/530/480 filters. It wouldn't be perfect, of course, since the real data from a
600nm filter could theoretically be anything.
True. This could be done using something like the Channel Mixer (Image/Adjustments/Channel Mixer), although the slider settings in there aren't the
least helpful to do it scientifically. It's also only linear, while in this case you'd want to apply it on a curve.
In theory, basically, you'd want to remove red from the red channel (darken it), and add some of green to it (to lighten it and bring it higher on
the spectrum). Like you said, it's not real data, however. It's an interpolation. Also knowing the specific frequency responses of the calibration
targets would be a big help to balance out the result.
All that said, I think that the images that NASA is putting up are a pretty decent approximation.