Combination of the infrared images of Sol 616 with R4 (red), R5 (green) and R6 (blue) give as a pseudocolor photo after some color enhancement an interesting result:

the blue-gray sharpe edged small stones (probably some vulcanic material) found strawn around the Erebus site look quite different in infrared but not in the visual band. There are at least three destinct infrared signatures giving probably a hint on their origin and chemical composition.

Here, on the left an R2 (red), R1 (blue)-picture is shown with synthetic green chanal according to the visual color taken from an L4L5L6 image and on the right the R4R5R6 picture.

The original Opportunity photos were normalized using a circular mask to reduce the radial loss of brightness to the edges.

This pair can be seen on the right (I am wondering that the thumbnails are not shown in the correct sequence).

The second image couple shows Bounce at Sol 68 in L4L5L6, L2L7 and R2R1 (from left to right with synthetic green in L2L7 and R2R1) for comparison: There is a possibility to make nearly natural looking color composites using the R2 and R1 channels.

This image triple is shown on the left.

(Image processing was done by collegue Mr. Norbert Gasch)

The interesting question is now: Does anyone have an idea what sorts of material this can be at the Erebus site?

Best wishes: Harald

You're using the JPG's from Erebus I assume - and that's a fundamental barrier in making any spectral claims. It can not be done. You can fudge them to make pretty pictures, you can measure how far things are apart, but you can NOT under any circumstances, using any magical technique, use the raw JPG's to make any claims as to the spectral properties - even in just a relative sense - of anything imaged.

Doug

Doug: I'm guessing you are saying that because we have no idea of the exposure times used for the various fiters, and will not know until the PDS release of these images. Is there something else going on in the conversion to jpeg?

Because the individual exposures are unknown, we sure can't say much about the spectral properties, but it seems that within the confines of this particular set of images we can detect relative differences between the spectral properties of the various pebbles.

Harald's image from Erebus is interesting in that it seems to show several different types of pebbles, where they all are very similar in color, looking at the visual bands. I previously thought we were seeing only two different kinds of pebbles; the blueberries and the dark cobbles. But this false color composite seems to strongly suggest several distinct mineralogies/lithologies.

Harald: A long time ago, before I understood that the individual filters were imaged with different exposures, I tried to determine the mineralogy of the blueberries by sampling their luminance in all the filters, correcting them according to the quantum efficiency curve for the pancam CCDs and their predicted exposure times, as published in the Bell paper, and then comparing the derived spectra to published spectra of various minerals. It was a lot of work, and my results were inconclusive. After the announcement that the blueberries were hematite I could see some crude similarities between my spectra and those released by JPL in an early press conference, but nothing one could be confident about.

Thanks for sharing that image with us. I have to imagine you are seeing a real difference between the pebbles in that image, but I think we would need additional information to identify the minerals. But to be honest, I am not an expert in spectral analysis.

I propose a simpler explanation. The sun changed angles enough that different surfaces of the angular pebbles are brightly illuminated in the different exposures.

Harkeppler: I've gone to sol 616 on the rover web site (not exploratorium) and downloaded all the images for the L and R cameras, and then did several color composites based on different band combinations.

For my composites, I combine bands in the order (RGB) output = bands (abc) in put, where I list the bands in order at the end of the file name, always using long-wavelength as red and short wavelength as blue. I tweek the contrast with "curves" in photoshop to improve contrast on bright surfaces and do an edge-filter sharpening in "LAB" on the "Lightness" image so as not to add noise to the color variations.

The L: 234 image composites infrared,far-red,red as R,G,B and shows little color contrast at long wavelengths in the Left camera. The L: 257 images uses near infrared, Green, and Violet filters as RGB and gives similar results to the more natural color 456 combination, but with more contrast. L: 567 composites Green, Blue and Violet as RGB and shows short-wavelength color variations.

R: 654 composites the 3 right camera infrared channels you said you used and should approximagely produce the same result. I simply do not see the wild color variations of some of the blueberries you've produced. Nor do I see color fringes at the edge of shadows in your image that would result from significant sun movement between filters. (Shadow fringes would be small as the height of the shadow-casting edges is small)

I'm wondering if you may have found a set of images taken at the same sun angle and with the same rover position but on DIFFERENT DAYS, and the wildly colored blueberries could have had dust added or subtracted by wind during the interval between pictures. Otherwise, I'm wondering if your image processing program might have color or brigtness "keyed" on something by accident and generated these bizarre colors. Your Bounce rock images look perfectly reasonable, so I don't understand what may have happened.

Yes - they all get stretched to hell and back. Basically, it's like the 'auto-levels' function in photoshop - and once you have an image, you have no means whatsoever of converting it back. Not to mention the conversion from 12 bit to 8 bit, and heavy JPG compression.

It's not just exposure and stretching, it's all the other calibration that goes on ( flatfield, dark field, exposure, temperature etc etc etc ) that we only get with the PDS releases as well.

Doug

The problem item is the stretching\auto levelling because we have absolutely no idea what the conversion was doing at the time. The use of the jpg format also affects the usability of the data, I suspect that we could compensate somewhat for it but I would not trust a spectral interpretation from an image that had been jpg'ed.

The other conversions could be compensated for to some degree:
8->12bit. We have the look up tables for these so that's straightforward.
Exposure. This is slightly tricky but with the regular sun shots and fairly precise knowledge of the light model at any point in time it is be possible to get a fairly accurate estimate of the exposure duration of an image.
Temperature. This is harder to deal with but once again it can be estimated albeit with a fairly high error margin based on already published data.
The Flatfield\darkcurrent removal and other calibration steps seem to be sufficiently well documented to allow amateurs to carry them out if they had unmodified raw image data.

IF we had temperature, exposure and (all of) the unmodified raw image data then we could do the rest. But since we don't have that it can't be done.

All we actually need is an l/p to the recent data on the workbook

Doug

Your words are strong and absolute. Some people are very bright.

Because the matter is absolute. Fact.

You could be the most intelligent person on earth, but you cant get from those JPG's to any sort of calibrated product. End of story.

Sorry to be so blunt about it - but there are too many people claiming to do 'real' colour images using the JPL JPG's online, and it's missleading and uninformed to do so.

Doug

It's worth pointing out that this is at least the third time this has been discussed here in some detail. I was at least partially the cause of a previous debate on this and very much appreciate Doug's patience with those of us who have caused the debate to be revived.

It would be nice if future missions provided completely raw image data in some lossless format in addition to the "stretched for easy viewing" variety but for now we have to be content with just working with the pictures.

Well - we do get the proper data, just 6 month in 'back pay' as it were.

I must admit - I'd like to see that time come down - but I also know that the scientists and engineers who do the amazing job they do, deserve the first bite of the apple

Doug

This whole discussion reminds me of the net kook who insisted he had developed proof for his "c" theories about a specific manned launch disaster (that occurred just short of 20 years ago, you'll pardon me if I don't invite unwanted attention by mentioning him, or giving other specifics) by determining the elements within the fireball via spectral analysis of a VIDEOTAPE that he had of the explosion.

Yep -- he put a crude spectrometer up against his TV screen and claimed that the light recorded on the video signal contained all the original wavelengths necessary to do a proper spectral analysis of the fireball.

Funny thing is, he made a big deal out of all the phosphorus he saw in his spectra... *giggle*...

-the other Doug

To be fair, I don't think Harkeppler or anyone else made an outrageous claim here. In spite of the jpeg stretching, the lack of calibrations, etc... it seems perfectly reasonable to expect that relative spectral differences could be interpreted as differences in mineralogy within this image.

What is most interesting now is the fact that edstrick's work did not reproduce anything similar. I'd like to better understand why that is the case.

Yep, gotta agree. Harald made a decent effort to check something out in a methodical way and I wouldn't want him to think we were a bunch of arrogant wags having a go at his efforts. It was a good question just unfortunately based on untrustworthy data.

Since I made the same erroneous assumption myself in the past I have to say it's not obvious at all that the image data is modified before it's put online. It's glaringly obvious as soon as you try to do anything serious with them.

The same IR composite image can be seen in this mars-news.de page.
That site is maintained by Holger Isenberg, and the IR composite is also credited there to Norbert Gasch.

A google search on Norbert Gasch shows Dr. Gasch really exists and is doing Martian image analysis, but I could not locate a site where he claims himself this IR composite is his work.

Puzzling .....

See - he makes this claim

"The next picture shows the same area in visible light as a human on Mars would see it himself."

From the JPGs. I dont know if he's misinformed, or intentionally misleading - he's certainly wrong in saying that Viking imagery is used as a calibration benchmark for MER data. It isnt. It wasnt with Pathfinder either.

Doug

I would even say it's pretty obvious when you just look at an image of the horizon and a 'second row' image (for a pan). The effect of the brightness levels in the sky causes another hue in particularly the blue (L6,7), due to the 'auto-leveling'.

There's so much you can do using whatever software package to make the scene look real based on what you know from calibrated output.

I like messing around with the .jpegs to get a good feel of the recent whereabouts of the rovers but must admit I spend more and more time working with the calibrated images since the amount of data is really getting huge now. We get the next release (sol 450-540) next Februari !

Nico

"Picture and discovery: Norbert Gasch "

LMFAO

Nico

I am perplexed.
I find articles by Dr. Norbert Gasch in sites as astronomie.de or www.extrasolar-planets.com

However I also found this one, talking about A natural "Stonehenge" on Mars?

And what is his affiliation, "Arbeitsgemeinschaft Raumfahrt und Astronomie", the "Working group for space travel and astronomy" ?

I followed your link, and then spent the next few minutes wondering why all my google searches were in Greek...

My German is limited to "Ich mochte ein stuck kasekuchen" so I cannot tell if he is a professor of whatever... but his few posts in English seem to be on the dramatic side - "WHY WOULD NASA CHANGE THE COLOR OF THE SKY" sort of stuff.....

Having just developed my underwater pics from a trip to Hawaii, I am keenly aware of the difference between "perceived color" and "real color" as if there were such a beast.

And now I will have another slice of cheesecake.

PS - It's obvious Opportunity has discovered ancient Martian Aquarium Gravel.

Aquarium gravel...

Nico

Hehe! "...Aquarium gravel..." If nothing else, it was worth having this little discussion just so Lyford could make that joke.

I finally found some time and just had to make my own R456 composite for the hell of it. It looked like edstrick's. I was about to call it quits when I noticed some slight coloration in the otherwise gray scene. I seriously cranked up the saturation, tweaked the hue knob back and forth, and ended up with a crude approximation of the original. I'm sure some of you who are more adept at image processing could make a closer approximation. Maybe this is what was meant by "color enhanced."

I don't have time to try to make a better approximation, so I am posting what I have now. The question now seems to be, can these slight variations be so radically enhanced to suggest the pebbles are really different? The answer to that is definitely beyond my expertise.

Yes, it seems that he has ramped up the contrast on those channels to make his point.... it almost looks like he colored each filter and then subtracted out the background to leave the colors more vibrant. I don't see how else you could end up with that wide a range in the same image without intentionally skewing the balance. Subtracting out the background and enhancing contrast seems okay to me for dust devil movies, but may not reflect a reality here.

As Doug remarked, attempting multi-spectral analysis with jpgs is non-starter.

My alarm always goes off when I read the phrase "as it would appear to a human standing on Mars," since that seems to inevitably segue into a NASA IS HIDNG BLUE SKIES ON MARS... But to give him the "benefit of the doubt," it would appear that English is not his native tongue, and perhaps some of his errors are just due to poor choice of wording.

Cosmic Rocker's saturation enhanced image does an excellent job of bringing out the spectral-response shading in the camera, probably due to varying sensativity of the CCD.

When I try to see if there is any useful color differences in the IR bands of the right camera, I generally grab bands 6, 4, and 2, compositing them as red, green, blue (long, middle and short wavelengths) Nominally, using Band 7 instead of 6, and maybe an average of bands 4 and 5 as the "middle" band (green channel) might give a longer "baseline" in the spectrum, much as bands 2, 5 and 7 in the left camera give stronger colors than the "natural color" bands 4,5,6. But when I generate bands 7, (4+5)/2,2 color from the right camera, the background color shading is much stronger than when I make band 6,4,2 composites. Apparently as the CCD's light response is dropping to near zero at the longest wavelengths, there is quite a bit more shading across the detector than in the other bands.

I'll sometimes make 2-color images from the right camera, using band 2 (near ir) and band 1 (blue or violet), using the average of the two bands as green: 2,(2+1)/2,1 color. These can probably be used pretty successfully with band 2,5,7 color images from the left camera as stereo pairs, but I'm stereoblind due to amblyopia (left eye corrects to 20:200)

I hate to say it, but I'm increasingly convinced that the aquarium-gravel colors in the image under discussion are entirely faked, either by accident or by intent.

Oops
My apologies, everytime I delete my cookies, I forget I am redirected to the greek google site automatically.

As to the IR composites, I would not be surprised if different kinds of pebbles at Olympia had different IR responses. However, this would have to come from calibrated pancams. And even if the difference is so huge that it is discernible in raw composites, I would at least demand the results were reproducible. Edstrick's and Cosmicrocker's attempts have clearly shown this is not the case.

Note that despite almost certain differences in absolute numbers of scene brightnesses in the different IR channels as sunlight amounts and camera sensativity decrease toward long wavelengths, and longer exposure times and wider filter pass bands (I think in channel 7) compensate, that the "auto-stretched" near-ir channel "color" images from the Right camera are essentially gray.

If (and I'm faking numbers here) the bottom 5% of an image gets contrast stretched to an 8-bit value of zero, and the top 2.5% of an image gets stretched to an 8-bit value of 255 (with everything outside those limits saturating black and white, respectively), regardless of absolute data values in the image, about the same relative levels end up black or white. Thus despite total loss of absolute brightness and zero-level information in the images, RELATIVE color information can be retained quite precisely.

And the relative color differences of materials in the near infrared channels are alway small. No infrared aquarium gravel on Mars that I can see.

In light of the mild feather ruffling induced by this topic, it seems that this would be a good item for a FAQ. Has there been any discussion about developing a FAQ for UMSF?

Given the amount of traffic here, I know it would be a huge task - but you don't have to climb the mountain in one day. Just the major "hot button" things that come up all the time...

Maybe some remarks are allowed:


The photos used are relatively new and no radiometric data are available at this time. So, everything I try to say is, that there is a relativ (!) but distinctive difference in the spectral reflectance.

to: edstrick

I have made a second attempt with R3 R4 R5 to document the process:

Left: after making a composite with R3 = red, R4 = green and R5 = blue

You can attribute other colors too, this is not a problem. The color diference can be seen here, I assume.

Right: after increasing only the saturation the things come out clearly. This picture can be processed further but I have not done here.

But have a look that most people can distinguish red hues better than blue ones.

to: tfisher:

>I propose a simpler explanation. The sun changed angles enough that different surfaces of the angular pebbles are brightly illuminated in the different exposures<

No, that is not the explanation because several pebbles look different. The sun angle effect should be the same on each.

to: djellison:

>You're using the JPG's from Erebus I assume - and that's a fundamental barrier in making any spectral claims. It can not be done. You can fudge them to make pretty pictures, you can measure how far things are apart, but you can NOT under any circumstances, using any magical technique, use the raw JPG's to make any claims as to the spectral properties - even in just a relative sense - of anything imaged<

Yes, Erebus Sol 616. The fundamental barrier is the result of the problem that JPL is handing out photos without radiometric data on time to bypass the freedom of information act and to reserve the complete data set to their scientist and their affilates. The freedom of informaion act only considered the photos theirself. Funny legal situation.

HK

Sorry. You are wrong. The different shapes and relief charactersitics of the pebbles will cause different faces to become brighter or darker as the sun angle changes. If this happens BETWEEN color layers, then the differences will manifest themselves as unusual colors.

tfisher's expalantion was spot on and I believe the very cause of this "anomaly".

No it's not - reread the explanations in this thread. We are trying to correct a serious mistake in the assumptions you are making about the images themselves. The images are modified before being posted and those modifications make it impossible to use them for your analysis. One of the sympotoms of analysing images like these in the way you have is that the results show large "unexplainable" variations. The actual explanation is that the images you are using have been individually transformed in order to enhance their visual appearance. That process is lossy and deforms the data in such a way as to render spectral analysis impossible.

The lack of parameters like exposure time, CCD temperature etc are not the main problem. Your analysis would have (some) validity if the images you were using were raw images from the CCD's but they are not.

I can't comment on the reason for this (it's probably just to make them look better for the average user) but the Freedom of Information act has nothing to do with it.

Nothing to do with FoI at all. It's bog-standard-science-mission-practice - release science data six months down the line so the people who made it happen get first dibs at it. That we're getting the quick-look-rough-and-ready-jpg's is an enormous bonus.

Doug

I personally think that the Good Doctor is straining at a gnat and trying to infer things that ain't. Over at the culinary forum down the hall ( "Kooks Online" ) they think he's on to something and might be using this technique to find fossil chicken lips in the meridiani plain...



--Bill

After thinking about this further, I think I've seen the light. It seems this IR false color composite is probably not showing anything significant.

I have a hard time believing this is a sun-angle thing, since the total ellapsed time between the images used is only 44 seconds.

What we are seeing are relative differences, and relatively small ones at that, confined to apparently random pebbles in the scene. Even before the major luminosity boost needed to create the aquarium gravel effect, some of these small variations can be seen in the IR composites.

You can also see some small color differerences in our familiar L456 composite in the visual band. Those pebbles that appear so dramatically different in the luminosity-boosted, IR false color composites are those that appear simply as different shades of blue in the false color visual band composite.

I didn't notice any of us getting excited about previous visual band images, saying "Oooh, look. There is a pebble that is not quite as blue as the one next to it." These only appear important because the false colors, arbitrarily chosen, are so dramatically different.

For me, the case is closed. It's too bad there would not be a significant TV audience interested in space forensics. I could get addicted to a series called CSI-Space.

In a way, we (I?) sort of do that with the L456 or L257 tricolor images we assemble. We use slight color differences to spot suggestions of compositional differences. As an example, Dilo has a color pan over in the Erebus Complex discussion. Some of the evaporite rocks have a purplish tint (which I'ev seen before), and in a couple of places, the ususal blueish hematite sand between paving slabs is rust-colored. I look at this and say 'hmmm, this is different', let's look closer.

FWIW.

--Bill

Exactly right, Doug. I'm involved in a large-scale experiment which has (with government approval) a two year moratorium on release of the raw data. The NASA/JPL guys are amazingly open about how much information they are willing to share immediately. It makes perfect sense to me to withhold a portion of the "best data". The people who have worked around the clock getting the mission to this point deserve all of the credit in the world, and certainly deserve first crack at their data.

If you want to contrast and compare, go over to the Mars Express website. You mostly just get PR blurbs on that site....

So what we are seeing here is just exaggerated differences. I could make any two pieces of granite in my backyard look different if I cranked the color gain up enough to get a little bit of difference, and then chose two wildly different false colors for each of the two slightly different colors. Maybe one of the rocks had only 5% more feldspar in it, but it would look like an entirely different breed of rock after performing my image exaggeration process.

It seems to me that people do this a lot with various images -- along with pattern matching of random shapes -- to infer just about anything they want.

Recently, I was trying to enhance the dark cliff face to see if I could find any evidence that it was really an outcrop of a darker material, or if it was just a shadow. As I "enhanced" and stretched it, I began to see weird little people in there! But it was all just an artifact of the process and my human-brain pattern-matching preferences.

Jpegs make things even worse because they add periodic fix-pattern noise that can be misinterpreted as real objects. It is like a constantly varying filter -- hard to tell what you have.

Fiddle with an image long enough and you can find/generate anything you want.

Scott

Exactly, Scott. It is wondrous that we can download image files taken through bandpass filters and easily combine them into 24-bit color images with readily available software on home computers. But the downside is that even "an absolute neophyte re. the theory of image processing" can do the same thing and with some time and little experience can arrive at wild conclusions. It's a perfect example of the computer principle of GIGO.

--Bill