Obtaining properly
balanced color results from the Pancam is a
fascinating operation which can be done on most common imaging softwares.
Color reproduction from multispectral filter greyscale images such as provided
by Pancam basically consists of obtaining the proper visible band images and
combining them to RGB, or Red, Green, and Blue as per their wavelength
values.
Given the band images provided are adjusted within the
proper brightness response curve for each of the three respective colors, your result
from simple RGB combination will be a "true" color image. Of course this
would be quite a rudimentary image, and normally in need of further
balancing, cosmetic, and aesthetic adjustments, but the components are all
present and significant further color enhancement is usually unnecessary,
in some extreme cases detrimental to image truity .
Now
we get to the Pancam, which is a remarkable piece of imaging technology, and
simply explained for being so complex. It has two "eyes", left and right.
They both "see" differently.. The left side sees almost entirely in visible
light, the right is almost exclusively infrared. Each is equipped with eight
different filters, which allow only certain wavelengths of the spectrum
to be seen in each of the images it captures. Here are the filter positions
by image/wavelength.
LEFT
CAMERA RIGHT CAMERA
L1. EMPTY
R1. 430 (SP) *
L2. 750 (20)
R2. 750 (20)
L3. 670 (20)
R3. 800 (20)
L4. 600 (20)
R4. 860 (25)
L5. 530 (20)
R5. 900 (25)
L6. 480 (25)
R6. 930 (30)
L7. 430 (SP)*
R7. 980 (LP)*
L8. 440 Solar ND
R8. 880 Solar ND
*SP means short pass filter, *LP means long pass filter.
One lens on each side is coated so it can observe the sun (solar ND)
L1 is clear, no filter
We
all know or should know that "true" RGB wavelengths are something that cannot
be nailed down to a definite value, the response is more of a curve.
There is no absolute "true" because color responses vary
depending on observer and other factors such as lighting and atmospheric
conditions.. The average human response for RGB is a given however, and
these values are..
Red 650nm
+/-,
Green 510nm+/-,
and
Blue
445nm-475nm
+/-,
If you need
references for that...
What Wavelength Goes With a Color?
http://webmineral.com/help/Reflectivity.shtml
http://hyperphysics.phyastr.gsu.edu/hbase/vision/cie.html#c2
We
can quickly see by referencing the filter wavelengths above in the left
column that none of the filters provided by Pancam are exactly equivalent to
the actual and normally applied RGB values, nor monitor colors. This being, it is not purely a "true
color" camera, even though the filters do fall absolutely near the accepted
values. The closest approximations to "true" RGB it natively supplies are..
Red L4 600nm (
or L3 670nm)
Green L5 530nm
Blue L6 480nm
( or L7 430)
Although
these bands are not precisely perfect color, beautiful close color images
without the overdriven reds can be produced from this data without much
further processing.
This can be
as easy as combining normalized bands L4, 5, and 6, reducing the blue and
green saturation levels accordingly (red will rarely need reduction, instead
it will need slight boosting if L4 is used for red), then
aesthetically cleaning and contrast balancing the image. In most of the
resulting products it
is necessary to adjust the uneven field illumination.
L4,5,and 6 color balanced calibration images,
Spirit and Opportunity.
Using this process will show us as
close to true images as possible from both rovers. To the right is an
Opportunity
calibration image. Notice the sky reflectors for both have a blue/grey/pink ruddy
tint.
We must note that as of yet we haven't a clue as to the
actual brightness levels in these
images, but naturally red is more intense than green which is more intense
than blue. Red covers a short range, green a middle intensity, but wider
spread, and blue the most diffuse, but wider still. The
web "raws" were stretched according to their wavelengths, or
normalized, so HSL reduction on the green and blue will be necessary to bring the colors into more proper adjustment.
This is required due to the nature of using multispectral imaging equipment
to get true colors.
Each filter only sees in the intensity of it's assigned
wavelength, therefore on 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. This is color
imaging, not rocket science, and the best balance is an aesthetics tuned
process that represents the calibration dial accurately.
NOTE: There are now plentiful
brightness calibration images being supplied for even better accuracy.
To sum this up
Proper techniques and
adjustments are the difference
between images that look like
this which is nice, but very ruddy, and
this. The
NASA/JPL color images we've been seeing and the first one linked above have been
made with the "wrong" red filters, the near infrared, reportedly done so in
order to better identify surface compositions. I would prefer to see the
sights as they really appear.. so would many others. Use of any other than
actual red green and blue filters will not produce an accurate color image.
Use of any other filters and then calling the result "true color" is
inaccurate and misleading. IR is not
visible red. Use of an IR filter for red is not accurate in close or "true
color" imaging, period. It's called pseudo-color when the proper
components are not present and one or more is manufactured from other
sources.
What it does is make for a paradigm fitting "pretty picture" with salmon skies
and too dark red surfaces. Brightness reduction for human response to ccd
imaging is not a secret, nor
is any of the color imaging done with the pancam
Here is a "pancam" picture taken on
earth. (No IR for red on this one)
Now before anyone goes running
off shouting conspiracy, trying to blame anyone or agency of dishonesty
(Especially the hard working MER personnel of whom I am colleague), or even dismissing this
issue as irrelevant, perhaps we should step back and take a look at the
bigger picture.
It
is most definitely relevant we have proper true color imaging of the Martian
surface, however, it takes a little time and work to apply the personal
beautifying touch to these images. Because the
bands are not quite RGB proper, it takes a little more than cursory
attention to attain the proper
color aspects. While I have poured a good bit of attention into meager sets of
full band data in order to make these Spirit and Opportunity color
images, they are at best a decent guess until we get full resolution images
and full ancillary data as it is to be delivered to the PDS. Take them as they
are.. Stunningly pretty pictures of Mars in color.
The MER team has a daily influx of not only photos but mission data. It is a
bonus that we get frequent new images, but it might be a bit unrealistic to
expect fully processed and calibrated totally accurate true color beauties
out of every downlink. This will happen gradually as the full record is
archived and made available. The world is clamoring for big fresh images on
top of this daily influx.. All that work, being done, yet to be done, and
still coming in to be done... Thus it is that the huge, undefined, hastily
constructed pans with the ruddy colors from using a "wrong" red, bad
stitch lines, and beige sky come to pass. I don't think they do justice to
the actual beauty of the Martian surface.
While there is no doubt in anyone's
mind that Mars is in fact a very ruddy color, at times it is very much
yellowish brown, earth-toned, and bluish skied. The differing ruddiness of
the images constructed from the proper color bands is directly dependent
upon ambient lighting conditions such as time of day, dust and water
cloudiness, distance and angle of surface being photographed to the sun and
rovers, materials being photographed, or any combination of such.
What is undoubtedly the most important
thing to remember is that no matter what the representations of "true color"
show, real red green and blue is...
Red 650nm+/-
Green 510nm+/-
Blue 440-475nm+/-
If it's not
that, or very near, then it's not "true" color
I offer here an alternative view, one done
with a fair degree of accuracy using common and measurable techniques.
One which also shows that Mars may be more than we ever imagined.
Two Worlds, One Sun.
I look forward to both Rovers continuing
massive success.. I also look forward to expanding
this study as new and more total data releases of this magnificent imagery
are acquired. Thanks for your interest, and keep checking back at my
collection for new images and accuracy improvements on the ones already
displayed.
Planetary color imaging, a high
form of both art and science..
-- Keith Laney --
For more insight into making sense of the images and the way Pancam captures
them, I recommend reading this article from "Above Top Secret", which is
fairly accurate as a whole despite the author's errant assertion that it is
acceptable to substitute an IR channel for red.
http://www.atsnn.com/story/30048.html
If you'd rather have an unbiased look at the actual scientific process
behind Mars color imaging, I suggest this article from Dr. Levin over
the above article
Solving the color calibration problem of Martian Lander images
Spirit Color Images
