Originally posted by ArchAngel
This may apply if the filters covered the entire spectrum and you were using CMYK to create the images, but they are created with RGB. It is the
ABSENCE of light in the range of the transmittance of the filter that shows color, not the other way around.
This makes me think that your understanding of hyperspectral imaging is not really what it needs to be, in order to discuss this properly.
First off, you need to understand how narrow-bandpass frequency filters work in conjunction with CCD cameras. From a perspective where that
understanding is in place, your statement makes no sense at all.
It can be easy to get confused if you come from an emulsion-film background and are used to the types of filters typically used with that technology
These are not the same kind of filters, because CCD technology is radically different from film. These are BANDPASS filters, which are the logical
opposite of putting a regular camera filter in front of a lens. These are also called interference filters and/or "dichroic" filters.
There's a really neat webpage that explains the difference between the filter types. Give it a view
Those who work with audio files are probably familiar with the difference between a bandpass and a band-reject filter, but many photographers may not
be. There are different terms to describe the behavior of a filter that BLOCKS a narrow range of frequencies... notch filter, band-reject filter,
bandstop filter, etc.
Pancam doesn't use the kind of filters that you seem to be thinking of. I can see where that could lead you to topsy-turvy conclusions, because you
might intuitively think that the L5 filter BLOCKS that range of frequencies, and hence that darkness indicates the PRESENCE of light.
That's not how they work. The narrow bandpass (interference) filters block EVERYTHING BUT that particular narrow range of frequencies... so the
presence of those frequencies gives a bright signal, not a dark one.
That's why there's no need for "negatives" when working with this technology... you don't have to take darkness and interpret it as brightness in
those frequencies. That makes the data much easier to work with, IMO.
Kano provided some links to nice reference materials in his rather canonical thread on the subject.
When a shot is taken on one of the Pancam cameras (either the left or right) with a narrow bandpass filter in place (say, L3 through L7), what happens
is that (basically) all frequencies of light that fall outside that particular narrow band are BLOCKED by the filter. The photons that make it
through are photons that are within that narrow bandpass (typically plus-or-minus 10 nm or so).
Those photons hit the CCD array, and create electrons trapped in the CCD wells, and the act of reading them out gives a rough "count" of how many
electrons were in each well for that particular shot. The more electrons, the stronger the signal.
That set of per-pixel signals is normalized to set the highest value at a reading of 255, and all of the other values are scaled accordingly. That
array of pixels is then processed, compressed and transmitted, and what we see posted is typically a JPEG-compressed version of that data set.
What does that mean? Well, the image that we see from an L5 shot, for instance, shows us bright signals only from photons that came in within a
frequency of roughly 520-540 nm. If there were very few photons in that range, the associated pixel will be dark.
If something is green in human vision when lit by sunlight, you can be well-assured that an L5 shot from the Pancam of that particular thing in
sunlight would show it as bright.
For something with chlorophyll, it would be bright in L5 while simultaneously dark in L3 and L6. So when you use a computer to mix Pancam shots of a
scene with L3 for red, L5 for green, and L6 for blue, you would see chlorophyll-rich plants as pretty vividly green on your computer screen.
And if I am wrong what colors can the cam not see?
I'll assume that when you say "colors", you mean colors that human eyes can register.
For something to be "color", but still show as consistently dark (i.e. not seen) on all Pancam shots with all lenses, the object in question would
need to have a remarkably strange response curve. It would have to absorb all frequencies except for a narrow-enough band that all of the response
curve would somehow fit between the filter ranges... but keep in mind that L1 is an unfiltered
shot. You see the response curve for L1 in the
background of one of the pictures you posted. Notice that the L1 position has non-zero response in the "gaps" that are not specifically covered by
individual bandpass filters.
Something with that radical of a response curve would almost certainly create an unusual situation when they compared L1 data to the other filters...
you could likely see it as somewhat bright in L1 but extremely dark in all of the other filters, which creates a "warning flag" immediately that
there is something exceedingly strange in the shot.
From that perspective, I have trouble imagining anything that a human could see, that couldn't be detected at all by the Pancam.
I would not have posted this if I had not already confirmed it with an expert on chlorophyll.
Yes, but your quote indicates that you're dealing with someone who is familiar with old-school photography, but almost certainly doesn't understand
hyperspectral imaging with narrow bandpass interference filters... which are the exact opposite of what typical photography buffs think of when you
say "filter". Dichroic (interference) filters are top-end technology, and give counterintuitive results for those used to gel and glass filters.
Of course that leads to inaccurate conclusions, and your statements about the absence of light indicating color. That's an appropriate mindset for
emulsion technology with gel/glass absorption filters, but not for hyperspectral imagery with narrow-bandpass filters. They work in exactly the
OPPOSITE way. Light means light, instead of darkness meaning light.
Before "standing by your conclusions" any more, please open your mind to the possibility that you have become horribly confused by the term
"filter" meaning the opposite of what you think it does (for this technology).
There's a really great technical PDF file
offered by Cornell that gives excellent data to help
understand how the Pancam really works. I'd recommend that it be read (and understood) by anyone who wants to talk intelligently about what the
Pancam "can't do".
It's not your father's Polaroid... it's radically different technology from that, and understanding the results and capabilities requires
understanding the difference.
[Edited on 2-11-2004 by BarryKearns]