Using the technique of putting color planes together as-is yields the blue color of the so-called blueberries. But is there everywhere a (relatively) blue coating on the surface? See this composite image:



Full Res 1024x1024

True Color Version

The blueberries are not blue - They are gray. Just assembling channels (like L456) without adjustment pretty much always yields images that are way too blue.

The PanCams on the rovers use different exposure times for each channel, to maximize detail. As a second step, the images are further normalized from their 12 bit internal storage to a full 0-255 8bit range, and finally compressed (lossless or lossy depending on quality needed) for transmission to earth.

So without knowing the A.) exposure time and B.) normalization range, it is mostly guesswork for us amateurs in trying to create "true" color images. JPL, on the other hand, has all this data available.

Here's one experiment I made in another thread: Guesstimate "true color" trench - (from this thread)

Nope, note the word "relatively." You throw your hands up a little too fast! Why do you even mention compression or mapping from 12 to 8 bits? These will just add a small amount of noise, not relevant to the effect observed.

The different exposure times (unknown) will have an effect on the absolute color (however that is defined), but the blueberries and the surface are *relatively* blue compared to other scene elements and that will be the case even if the images are normalized as you suggest.

> Mark Lemmon: "We've looked at the floor of the crater itself and it's
> exciting. In particular, this area is covered by finescale sandgrains and
> these irregular grains coming down from the outcrop potentially and the most
> spectacular are these rounded spherules being called blueberries because
> they're relatively bluer than their surroundings." (Thursday, Feb. 12)


So the effect shown above is real. And the surface is relatively bluer than the subsurface. Why? It is evidently a very thin coating.

Perhaps I'm a little trigger happy, yes, but I guess I misunderstood what the point of your post/question was then. Of course they are more relatively blue.

The point about the 12 to 8 bit mapping/normalization is that the brightness values in the raw images are not necessaily proportional to the other channels. (0 might mean a real brightness of 20 and 255 might mean a real brightness of 30 for one channel, but in another it could be 0->0, 255->10000)

As for the whole surface having a relatively blue coating, it doesn't seem that way to me at least. The spherules and some darker rocks appear to be the only source of relative blue-ness. In JPL released images, the rest of the surface appears to be more typical red. (See image here from "charlie flats": http://marsrovers.jpl.nasa.gov/gallery/pre...lats-B033R1.jpg)

> As for the whole surface having a relatively blue coating, it doesn't seem that way to me at least.



All the pixels within the undisturbed area are predominant in the blue in this image. Including the spaces between pebbles. This contrasts with the disturbed soil, which is red-violet. How does that not infer that the surface there is relatively bluer?

Again, the question is not one of absolute color (I linked above to a true color image where the effect is not evident). Indeed, false color images are better for discerning differences in materials on Mars. Not to say a difference in color so obtained necessarily means a difference in material, there might be other subtle factors involved. But in this case the blue does seems to able to be smooshed around by the trenching activity, like it really is a patina of some kind.



PS: The image you linked to above doesn't have a whole lot of blue in it at all. Almost no points, including the spherules, are predominant in the blue. It is in fact a true color image such that the spherules come out gray. I'm sure that if the soil was scratched there, it would show up very red, and the surface would be relatively bluer in comparison.

PPS: These images are a composite of L2, L3, L4, L5, L6 & L7.

You are corecct, when it comes to relative blue-ness, soil underneath < top soil < spherules.

I suppose the top soil might be more gray because of being affected by the sun, or because there is some fine gray dust on top. That dust could have eroded of the spherules, or been blown in from another source.

Perhaps they should try the RAT-brush on the top soil somewhere?

Thing is, it is very thin and fairly uniform, so I would lean towards the sun effect rather than windblown dust.

Perhaps it's a thin crust of that hematite they are (or at least were) so eager to find. Does the reletively bluer color extend to the rippled deposits?

From your above post:

The different exposure times (unknown) will have an effect on the absolute color (however that is defined), but the blueberries and the surface are *relatively* blue compared to other scene elements and that will be the case even if the images are normalized as you suggest.




This is not entirely true, the PDS headers are included as timestamp files along with the maestro dataset. This data includes a field called "INSTRUMENT_STATE_PARMS.EXPOSURE_DURATION". This field shows a MASSIVE difference in exposure times, trending longer as the Left filter number gets higher.

I just finished plowing through newest maestro datasets spirit 5,6 and opportunity 4, (as well as the other 6) to get these relative exposure times, normalized to the shortest exposure (L2):

L2 : 1.00
L3 : 1.86
L4 : 2.13
L5 : 5.25
L6 : 7.50
L7 : 32.22

and for Spirit

L2 : 1.00
L3 : 1.57
L4 : 1.77
L5 : 5.22
L6 : 7.81
L7 : 26.47

(i should say upfront, there is a huge amount of variation in the exposure time ratios, thus the L7 in one sequence might very well be twice as exposed as the L7 in the next sequence...this is done by autoexposure, so a legitimately less blue site will have much longer exposure in the blue, to record what detail *is* present)


The above values are partially explainable as a compensation for the differences in the Quantum Effeciency of the CCD itself, as it has less effectiveness at the blue end of the spectrum than the red, but even compensating for this (using the R, or absolute responsivity, values as given in the Bell paper describing Pancam activities, for each filter and the bandpass values of each filter, to determine how "wide" a spectrum of light is being allowed through the filter), one still ends up with relative brightness values that say an L2 pic is about 5.4 times as bright as an L7 pic. This means that any sort of image-merge to color channel information needs to account for this relative brightness, thus toning down in large part, many of the blue features.

Also, to keep in mind, the exposure times on the L7 images is sometimes as long as 10 seconds, making the issue of CCD background noise a much greater concern. The "flatfield" behind any usefull information the image might contain is much greater for L7 pics, thus giving them even more "lightness" than they deserve.

JPL itself will release, as part of the PDS dataset, all the images as fully corrected, "reflectance" images, which should be perfect for image merging, since they would have an absolute frame of reference for brightness.


also...in reading the Bell paper, i realize there is a perfect resource to describe the activity of calibrating the images, though I can't find a way to get the actual document. If anyone knows where one could find a publically available copy of :

Bell, J.F., III et al., Mars Exploration Rover (MER) Project Pancam and Engineering Cameras
Calibration Report, JPL Document D-19826, MER-420-6-700, Jet Propulsion Laboratory,
Pasadena, CA, in preparation, 2003b.

I'd be much abliged to learn of it.


Daniel Crotty

Are you referring to this 98 page document?

http://europa.la.asu.edu:8585/PGG/greeley/...f/bell_2003.pdf
(LOTS of info about the cameras)

No, though that paper is a huge resource in and of itself (that is where the R (absolute responsivity) values i quoted above were obtained. I'm looking for the 8th reference listed in the bibliography of that paper. The calibration report itself was not completed at publication of the paper you linked.