First off, thanks to Kano for the truly yeoman's work done on collecting this excellent information all into one spot. I've been more than a bit
bothered since seeing the January 8th mosaic image, and this info at least allows me to explain
why the images look the way that they do.
However, I still have several problems with this approach, and more than a few unanswered pointed questions.
You wrote:
But, do not despair!. There are still a few which we can get a very close approximation of the actual colors. For example any picture that has the
sundial or that pole (or any other white part of the rover) visible. These images we can be sure are close to the true-color images, with the only
difference that any overall red tint will be lost.
and
So all we are able to do so far is show that the sky and ground color we have been seeing in the released NASA images have been accurate. As you would
expect really.
I would say that we can't really go that far... having the low-gain antenna in a ground picture allows us to (at best) compensate for the
normalization process. But it can never allow us to recover the "contaminated" red channel.
What do I mean by contaminated? Basically, that we cannot recover the set of colors that humans (and more constrained, the computer RGB display
system) can see and work with from the L2 data. L2 is simply inappropriate when trying to construct "human/computer friendly" color images.
Let's take the blue paint chip (and the insulation, and the NASA logo, and the flag, and the Columbia Memorial... anyone else thinking that using a
high-IR response pigment in ALL OF THESE was a bad idea?).
Since the blue has such a strong IR response, we end up with red channel pixels that are bright because they have a strong-IR-response blue in them
(and green to a lesser extent). We can also have bright red channel pixels from objects that actually ARE red in the human/computer range.
The problem is, after the image is produced, there's no way to say for certain why a particular random pixel is bright in the red channel... is it
bright because it is bright in the blue, or bright in infrared?
That means that there is no set of color adjustments to an L2-L5-L6 image that can consistently and reliably produce something close to what a human
would see standing there.
Further, Kano wrote:
So even if there was something blue/green on the surface that was extra bright in the near-IR range (very unlikely really) it would still be noticed.
You see, this is part of what bothers me about this whole thing... if we are approaching this scientifically, then we shouldn't be applying our own
personal biases about how "likely" something is to exist on Mars. We should use the tools that are available, and let the evidence tell us what is
there...
Bias and prejudgement of the outcome is what justifies people using an inappropriate subset of the available tools, and then explaining it away after
the fact. To wit:
The images shown by NASA are as close to the actual appearance from the surface as they can get. The colors are as true as fifteen million dollars
worth of camera and image processing software can get them. As accurate as any digital image can be.
I contend that this is, quite simply, wrong. The constraints on how close to a "true" image we are getting are
a direct result of their choice
of filters!
And that's the heart of the real problem that I'm having... there are filters available that allow for the display of something much closer to
"true human color" than L2-L5-L6. Namely, the inclusion of the L4 filter.
The L4 filter does not carry any of the erroneous IR data and attempt to map it on to the human / computer "red" channel.
NASA/JPL/Cornell have taken (and constructed) a variety of pictures using the L4-L5-L6 comparison, and have generally done a pretty good job of
representing what a human would see.
Examples:
The signed plaque, including the US flag and NASA logo:
2P126632659EFF0200P2899L4M1.jpg
2P126632703EFF0200P2899L5M1.jpg
2P126632747EFF0200P2899L6M1.jpg
The DVD with Lego character Astrobot Biff Starlight:
2P126556727EFF0200P2205L4M1.jpg
2P126556872EDN0200P2205L5M1.jpg
2P126556928EDN0200P2205L6M1.jpg
Pictures of the Pancam Calibration Tool:
2P126802413ESF0200P2110L4M1.jpg
2P126802302ESF0200P2110L5M1.jpg
2P126802332ESF0200P2110L6M1.jpg
2P126802413ESF0200P2110L4M1.jpg
2P126802302ESF0200P2110L5M1.jpg
2P126802332ESF0200P2110L6M1.jpg
OK, let's walk through a reasonable strategy for producing "human friendly" color images of an unknown landscape, given the tools that we have.
There is a color calibration tool mounted near the edge of the Rover. The following frames show that it is not difficult to shoot a full-sized (EFF)
or even downsampled (EDN) image which contains the calibration tool and a
substantial amount of terrain in the same shot:
2P126824507EFF0200P2303L2M1.jpg
2P126824540EDN0200P2303L5M1.jpg
2P126824562EDN0200P2303L6M1.jpg
The placement of the tool close to the edge of the Rover is, of course, not an accident. It is there precisely so that the terrain colors can be
calibrated against a known quantity.
The natural place to start when trying to determine which filters should be included for construction of a color panorama for human consumption...
would be to take a shot using this layout with each of the different filters.
One of the aspects about all of this that I find a bit unsettling is that there appears to be
no shots whatsoever using L4-L5-L6 which contain
both the calibration tool and the terrain... despite the fact that this is the obvious shot for calibrating "near human" images.
The vast majority of the calibration tool shots sent back are ESF shots, meaning that they have been "sub-framed" at the Rover before
transmission... they have been cropped.
I can see a bandwidth argument for the majority of the calibration tool shots... but ZERO shots of the terrain with the calibration tool using L4? I
find that a little hard to swallow.
They were willing to spend full-frame bandwidth for the picture of the plaque, using three full EFF frames. Even for the Lego character, they spent a
full EFF L4 frame plus two downsampled frames.
But we can't have even one EFF or EDN L4 frame (along with the L5 and L6 already taken for the color images) which has the terrain and the tool in
the same frame?
I don't buy any of the ridiculous consipacy theories about the Rovers not being on Mars... I'm certain they are there. I don't even believe that
NASA is doing any unsavory tampering with the data.
But I find it more than a little curious that the easy, obvious picture to take for proper landscape calibration in human-friendly pictures just
happens to be always overlooked.
It wasn't overlooked for the plaque... it wasn't overlooked for the little Lego Astrobot either. But when it comes to giving us an image which
allows us to calibrate the
landscape in human-friendly colors, we always seem to come up empty.
That set of choices fuels speculation, and they are bringing it on themselves. So I'm going to speculate...
Why send L4-L5-L6 data at full frame sizes for the plaque and the DVD (which people already KNOW the colors of), but never L4-L5-L6 in full frames for
the calibration tool?
The DVD and Lego shots can both be easily framed at 1024x1024 without showing
any of the surrounding terrain... and are therefore "safe"
shots to show in L4-L5-L6.
If the calibration tool is shown in L4-L5-L6 at full frame, the panorama makes it obvious that there is plenty of room for the terrain in the shot...
so either the real terrain colors can be calibrated, or they would be forced to choose a framing that makes it obvious that they are AVOIDING showing
the terrain.
One last disturbing bit that I've found: The "Maestro" software package allows you to see a very minor subset of the frames from the Rover Pancam.
One of those is a picture of the Columbia Memorial, which shows in a VERY nice blend of colors... red, yellow, green, dark and light blues.
For those of you who don't have Maestro, I saved a copy of the picture at
ebloot.com...
In the Sol002 raw data archive, you can see the L4 and L6 images, but the L5 was never released (as far as I can tell). So where did Maestro get the
green channel data from?
If they didn't publicly release that frame, might they
already have taken the L4 frames needed, and simply chose to not publish them (making a
human-friendly color balancing effectively impossible)?