PDA

View Full Version : Pancam images and RGB channels



brownpau
2004-May-04, 04:20 PM
Jumping off from this thread (http://badastronomy.com/phpBB/viewtopic.php?t=10643), I have a question for the Photoshop/GIMP people here: which pancam filters should I be using to generate quick color images via copy-and-paste into the RGB channels?

Given the filename key (http://marsrovers.nasa.gov/gallery/edr_filename_key.html), I've been estimating L3, L5, L7 || R3, R2, R1 should go into the R, G, and B channels respectively, but the results aren't always optimal -- I often get a B/W image, or else something with lots of washed out bright blue. I'm not sure if I'm getting my channels mixed up because the reverse order of the Pancam R filters is throwing me off, or if my wavelength estimates are wrong, or if Mars is just really weird-colored. How do the JPL pros do it?

Squink
2004-May-04, 05:18 PM
It's been a long time since I've tried constructing color images from greyscale, but how are you weighting the different channels? Just adding them together gives poor results. This person (http://www.radiance-online.org/pipermail/radiance-general/2004-March/001545.html) uses luminance = 179 * (0.265*R + 0.670*G + 0.065*B), which is about how I remember it. I'm not sure how the people at JPL do it, but Nirgal's version of the crater (http://www.badastronomy.com/phpBB/viewtopic.php?sid=a3cf56c678526f22c7a3166529bbe4be &p=253631) looks to emphasize reds and blues more than Space Daily's (http://www.spacedaily.com/news/mars-mers-04zzzzg.html) sepia print imitation.

brownpau
2004-May-04, 06:35 PM
but Nirgal's version of the crater (http://www.badastronomy.com/phpBB/viewtopic.php?sid=a3cf56c678526f22c7a3166529bbe4be &p=253631) looks to emphasize reds and blues more than Space Daily's (http://www.spacedaily.com/news/mars-mers-04zzzzg.html) sepia print imitation.
Heh, SpaceDaily's looks like a simple colorized B/W navcam image, which is what I'm doing with my Mars Wiggles. Nirgal says he's working with navcam too, but whatever he's doing, it's utterly awesome.

Nirgal
2004-May-04, 06:36 PM
unfortunately it's not possible to reconstruct the true color images
from the multi-channel pictures published by JPL :(

this is because JPL "normalizes" the exposure of each channel individually
(in order to get the full range of greyscale values in each single channel)
While this makes sense if one considers each channel índividually as a single
greysacle image it unfortunately destroys the "color proportions" among the channels if we want to put them together to form a composite color RGB picture :-(

For example: suppose the *real* color composition of a peace of martian surface rock (fo simplicity lets assume just one pixel) was 70% red 20% green and 10% blue.

Now what the JPL camera does is to compensate for the comparatively dark blue
channel (either by using alonger exposure time or later during a post-processing digital corection) So the result woul be, say,
50% red 60% green and 50% blue .... clearly the blue would be over-emphasized in the composite RGB image

The sad truth is that without the information of the exact exposure/normalization/contrast-stretch etc. etc. that JPL uses on each
individual filter channel I'm afraid that there is no way to do accurate
composite color images.

NAS/JPL: *please* also publish those additional information with each raw Pancam-Image !

jumpjack
2004-May-04, 07:13 PM
I eventually discovered a nice, little program (only 664KB download) to play with NASA raw images and build my own "true color" images:
Image analyzer http://meesoft.logicnet.dk/

Just load the three RGB components, then choose Image/CombineImages from menu, and that's all! :D

VERY approximately:
L2 INFRARED
L3 RED
L4 ORANGE
L5 GREEN
L6 BLUE
L7 BLUE
So, you can use several different combinations to obtain near-true color images; for example, L3,L5,L7 , or L4,L5,L6, or L4,L5,L7 and so on: wou'll always need L5 in green channel, then you can play with R and B components.

All R filter are INFRARED, even more than L2, so they are not suitable for true-color imaging.

Details about rovers' pancams:
http://www.highmars.org/niac/education/mer/mer00.html

Happy imaging! :D


Luca

Nirgal
2004-May-04, 10:11 PM
the problem is that even if the rover cam's filter wavelengths would
exactly match the "true" Red,Green,Blue channles, there is still the more serious "per channel exposure normalization" problem I mentioned above.

Actually, assigning the proper pancam-filter wavelengths to red,green and blue channels ist not the main problem (the L3,L5,L6 would be close enough for a very good approximation ;)

The real problem is the fact that each pancam channel was shot with
a different exposure time in order to utilize the full range of greylevels for each channel. (otherwise the blue channel, for example, would be much darker overall the image than the red channel ... but the JPL raw images
have approximately the same overall-brightness level.
So, unfortunately, we can not simply combine the cahnnels and get
an approximate true color image.... unless JPL decides to
publish not only the filter wavelengths but also, with each raw image,
the exact exposure time.

jumpjack
2004-May-05, 07:35 AM
Image analyzer http://meesoft.logicnet.dk/

Just load the three RGB components, then choose Image/CombineImages from menu, and that's all! :D

...ooops! #-o Don't forget to onvert the three images to grayscale by pressing CTRL+G, else you won't have anything to combine!



Details about rovers' pancams:
http://www.highmars.org/niac/education/mer/mer00.html

I just discovered reading these pages (and also Nirgal reply to my post) about the different exposure values available for pancam images: really annoying... :evil:

Luca

Nirgal
2004-May-05, 11:43 AM
Keith Laney has the best reference on the web about calibrating MER color pictures (the quality of the pictures on his site speaks of its own):

http://www.keithlaney.com/spirit_color_images_calibration.htm

he addresses the problem of "exposure normalization" in more detail:


We must note that as of yet we haven't a clue as to the actual brightness levels in these images

and suggest the following estiamted "workaround" to correct the channels
for the "exposure normalization"


RGB combination each color channel is maximized, needing reduction according to visual response curves. A round estimate for this reduction falls near the following RGB ratio; 1,-3,-6. A strong amount of color imaging "knack" is also quite helpful.

very interesting and worth reading article :)

Now I can't wait for the whole set of Filters for the Endurance
Pancam images (so far we seem to have only L2(infra-red), L5(green) and L6(blue) ...
one of the "red" channels L3,L4 is still missing ..