Timeline, Testimony, Processing, and Politics of a Bad Themis Image

By Keith Laney

 A bad image?

I first got involved with the Themis IR Cydonia image when it was the hot topic at The Enterprise Mission Conference on the late evening of July 24th. A poster known as Aramus had started a threaded discussion concerning this new image.


 I looked at it for a few minutes via the link provided, and after cursory review pretty instantly came back with a hearty "IT SUCKS!" post, which was resonated coast to coast by Richard on the Art Bell show. The reason for this was a dull looking low resolution B/W that at the time simply turned my stomach. This was supposed to be our "promised" precious Cydonia IR image. I was furious! No colors, 9 small photos composited on a large strip. Not at all like the MOC images I process. I didn't even bother to download it. I had seen long time friend Holger Isenberg's first attempt at a composite version, and decided I'd rather continue doing MOCs.

 The next evening while doing my usual (surfing through the conference replying to this and that message) sometime shortly after 10:00PM on July 25th, I chose to reread the "New Cydonia Image is UP!" thread again and catch up with what others thought about it. This time, when I came to the link posted to the Themis page, I decided to look at it again, just to see if I might have been a bit harsh. I clicked the link and got the Themis page containing the image. This time I downloaded it, and in tiff version (of course), because of the high quality of that particular compressible lossless format. After opening this up in my favorite imaging program, I noticed that it was very "blocky", perhaps even what can be described as "foggy" looking,

 I tried a couple rudimentary cleaning processes - but then I saw that I was doing nothing really to improve the image- so I just closed and left it alone. I hadn't studied much on multispectral or IR images at that point, but did understand the basic concepts behind color combined images and layer registration. I came to understand how each of these images are but representations of IR bands captured by the different filters on Themis, and that these bands could be combined and colored to represent their spectral bands in composited images.

 Being curious, wondering why the wonderful Themis team didn't give us a color IR image, the next step I tried was to assign these different images to different colors and combine them using standard RGB layering techniques. Of course, this is not accurate processing, nevertheless they were colorful, with the colors varying widely across the image.

On July 29, While still trying to figure out the proper method for preparing these images, I noticed a post on the message board by Mac Tonnies in which he stated that these images are scientifically valueless.


He was replied to by a relatively newer member named Bamf (better known now as Noel Gorelick of ASU), Who said this,

"Quite clearly you don't understand what you have, or even what the term "scientific" means. From this one image, it's possible to determine surface temperature, thermal inertia, rock abundance, grain size, reflectivity and approximate chemical composition for every feature in the image. It also maps any ground water or ice, and provides atmospheric dust opacity.

Just because you've failed to see anything in the pretty pictures, doesn't mean the data isn't scientifically valid, valuable and useful."

Then, a post later he says

"These images deliver both the CHEMICAL COMPOSITION and THERMAL PROPERTIES of the area. From this data, it is easy to show that the face is made out of the same stuff as its surroundings. It is not made of metal or plastic and there's nothing even slightly abnormal about the composition of the area.
The hope that the night-time data will produce different results is fantasy."

 I wasn't willing to take his word for it, especially as I did not know that this was the same fellow that created the image we were looking at. I had to challenge his assertion that there was nothing abnormal about Cydonia. We've all seen the images, and there is an extreme amount of geologic and mineral diversity all over that plain! Not only that, but this diversity has been explored and documented already.

A 1998 paper by James L. Erjavec of the Society for Planetary SETI research finds that

"Studies of the Cydonia Viking images (Erjavec, 1994, Erjavec and Nicks, 1997), showed that the alignment of pedestal craters, scarps and landform aprons, as well as the distribution of impact craters, are the result of a number of geologic processes and that the Cydonia area experienced several episodes of erosion and exhibits a surface expression that is polygenetic in origin. The curvilinear signatures of a series of scarps and ridges south and southeast of the landform called "The Face" resemble those of shoreline and strand line features on Earth. In addition, a bimodal impact crater distribution between the knobby terrain to the west and the cratered plain to the east, leads to the inference that at least two types of geologic influences were at work in forming the present landscape. The above surface features in combination with evidence of volcanic processes, fracturing, outgassing(?), liquefaction(?), slumping and faulting, indicate that this area of Cydonia is geologically complex and may be representative of a dynamic Martian system. Moreover, the simple erosional models previously suggested for Cydonia, which relied on aeolian or fluvial processes and differential erosion as land forming mechanisms, do not adequately explain the complex interrelationships noted in the Viking images. "

 I then set about in earnest to learn how to process these IR's, even going so far as to post publicly one of the invalidly done yet colorful band workups I had already played with. Only later did I figure out that I had just challenged the very fellow that was more than likely responsible for the image I've got!

 Now of course I was challenged back and derided a bit for my "insolence", From the read of it at the discussion board it does appear that I was goaded into learning the actual process, but no matter, the actual point was to learn to process these images. I already had testimony from the maker that the image is valuable, and his review of what the image would show. As well, the text on the Themis image page states that the image in itself is valid. If there was something to be found in it, I figured it would have to be one of us to do it. This was especially applicable since the Odyssey team had been unforthcoming in showing us the false color multispectrals made with the image bands.

 The Processing.

Seeing how band rationing and decorrolation is quite an in depth study, I went on a netwide jaunt to collect all the image processing info I could find. After learning more about the process, I saw that the software I had to do this with was hugely inadequate. This led me to a company called Research Systems Inc, of Boulder, Colorado. <http://www.researchsystems.com>

 They are the maker of a wonderful software package called ENVI 3.5 and IDL. This thing is highly touted as THE commercial multispectral imaging tool. Using this software, what was a long hard road to multispectral imaging turned into a simple run through the tutorials and inputting band images. Anyone can follow its directions, and ENVI provides point and click usability with an ease I find astonishing. Needless to say, I highly recommend it. It will absolutely FLY! Given that we are due for a major release of Odyssey data soon, it is worth obtaining.

 I took my originally downloaded image and started running it through the software, following the guidelines for proper processing found in it's tutorials.

Here are the general steps used to achieve my results with ENVI.

 Due to the one large composite image method used by ASU in releasing this 8 band image, all the individual IR bands were by necessity first cropped from the parent. These were then rotated to vertical, and rudimentarily cosmetically cleaned by a blanket optimization process using ENVI approved and recommended filtering methods. They were then carefully registered to properly overlay. This is necessary because the separate images do not fit precisely over each other, tending to dither either left, right, up and down by varying degrees and combinations. The image size of each of these separate bands when rotated left 7.5 degrees turns out to be very approximately some figures we might have heard of before, 333 X 1947!

Already we have a conundrum.

 It is flatly stated that the image is 32x199.1km at 100 meters per pixel resolution. The individual images should by all means be 320x1991 pixels! This is of note, as the images on the strip are approximately 333x1947 (given cropping accuracy tolerance within a pixel or two within each image).

 After registration, ratio, and processing, the images were then opened in ENVI and greyscale divided in various band combination pairs to create ratio images. Given that there are eight individual IR band images, 6.62, 7.88, 8.56, 9.30, 10.11, 11.03, 11.78, and 12.58, this makes for an interesting lot of available combinations. (7.88/8.56, 10.11/11.03, 11.78/12.58 etc.)

 These ratio IR band images are then taken in various combinations and assigned color, in this case usually in preference to their wavelengths, blue being short, green mid, and red high. For instance, ratio image 6.22/7.88 would be assigned to blue, 9.30/10.11 green, and 11.78/12.58 red. They are then combined into RGB false color multispectral images. The resulting image is then fit for spectral analysis, and further study for thermal and compositional properties. This is attained by equalizing and classifying the image further to create hyperspectrals.

 Color multispectral images can lack referencial surface detail and resolution. We can "add" it back by registering and applying the multispectral image as a color layer onto a visible or "luminance" image. The higher the quality of visible image used, the better the resulting image. In this study I chose Themis Visible Cydonia image 20020413 for a whole image & high res D+M overlay, and MGS MOC image E03-00824 for overlay of the "Face". Because of this technique it is possible to apply the multispectral color data onto a wider variety of corresponding visible images, even those from other sources, including the Viking Cydonia images. Images done in this manner display the intricate surface details of a visible image while colorfully displaying the false color multispectral properties, making for both an asthetically pleasing and scientifically valuable product.

The evaluation and comparison

 The bulk of the image processing done with the original image was completed between July 29 and Aug 30th.

This was work done on the only version of the image I knew of at this time.

 I was absolutely ecstatic and taken back during this period by the results of running these band ratio images through ENVI. So much so that I went back to Enterprise Mission BBS and changed my "This image sucks!" post with an apology. The color variations on the plain multispectrals are spectacular. The overlays are so color rich and clear as to be phenomenal. There are linear shaped blocks of various shades in thermal IR color representation all over the image, especially where there is what appears to be sedimentary deposition. They are not visible on exposed areas or the monuments. I decided to ask the guy who made the image... Noel Gorelick aka "Bamf"

"So, you've found out the secret" Noel replied when I asked him about them. Then this was explained by him later as basically an IR imaging dither caused by the temperature drift of the instrument and differences in spacecraft and filtering position while recording the targeted area sequentially. Dan Smythe (whoever he is) had also affirmed to me previously that the image had a "Waffling" and we would likely find a "honeycomb" pattern

Here is a screen shot of this conversation, on Aug.10th

 After perusal of mine and the other (2) color IR images so far put out by ASU, I see no way that this description could be entirely accurate, the blocks cannot be replicated or totally removed without seriously altering the image, which would of course make it useless for spectral evaluation. The same effect is not at all on the other images either. Regardless, Noel had confirmed to me that these blocks were on the original image, and later that they were looking for some way to remove them. He even wanted to know what I use on the MOC images I processed for Marsoweb. I have to ask... What, did Themis just suddenly decide to trip out over Cydonia? Why not over Marineris?

And about calibration... doesn't and hasn't the Themis webpage the image resides on always said this?

"The calibrated and geometrically projected data from all of the nine surface-viewing filters are shown in this figure".

The politics

  I had been sharing my imaging results with many others on the net in various message boards, and after preliminary review of them many were astounded also. The strange thing about it was that no one else was getting near the same results as I. Holger Isenberg wasn't, Steve Wingate wasn't, (In fact he only seemed interesting in debunking this image, which sadly he has looked very foolish doing)  Richard Hoagland wasn't.. No one that tried did. The best attempts I saw from others were rather streaked and the color values were smudged, weak, with high noise levels. Not very good examples of true false color multispectrals worthy of the capabilities of the wonderful Themis, which for the most part are supposed to be clean and colorful where there are thermal and compositional differences being imaged.

 The reason for this became apparent when I went back to the Themis site on August 26th. I found that the image now there is very different than the one I downloaded on July 25th. I could not believe it. This image was much "prettier". This was obviously not the same one that Noel just described as having blocks. In comparison, it looks very much like a visible image. This was the reason why the others were not getting the same results. I for some reason had a different image. Somehow, This is strange because the only source for that image on July 25th would have been the Themis site. No doubt my image came from there. This is further confirmed by comparing the official present image with what Noel described to me above. The block effect is not noticeable on the ratio images made from the "official" image, it is not evident in the individual band images from it either, yet it is in both on the image that I got, just as described.

 These "blocks" are in no way a result of any further processing that I have done on the parent image, these were on the image the day I got it. I have done slight processing work on my original dated file download image and saved it. Nothing destructive, and for the simple reason of trying to lessen the severity of the blocks while comparing it to the different clear image from the Themis site at present. This was on Aug 26th as evidenced by the modified file date. Original copies were made and saved from it beforehand, and all the work and ratios I have for the color multispectral images were made from it beforehand as well. I did this quite inadvertantly, but because the processes done on it are necessary and compatible for what processes are being done to the image to make quality multispectrals that is of little consequence. A lot of people have done a lot of torture processes on the image to try to discredit it's pedigree, but as of yet no one including myself has been able to replicate it using the image that is at the Themis site now.

 I have so far processed nearly every possible band ratio combination from my downloaded image, the "official" Themis page image of the present, and the other separate version from an associate obtained independently that is very similar to the one I downloaded. Without exception, and in every band combination tried, the version I got on July 25th is superior both in color data levels and clarity.

 The Official image at the Themis site appears to be a prettied up, heavily destreaked, and warp registered later version of the image I recieved. This is extremely important, because my image has no hints of destreaking or warp registration. Both of these are irreversible processes. There is no way I could have made my image from the version now at the Themis website. Not even had I wanted to.

 The inferior looking image produces superior multispectrals. That's all there is to it through the eyes of this processor. The official IOD makes a good resolution layer.

 The image data also confirms somewhat that this image is "different". Clearly in the header of my image, (Which is a tiff) and unalterable by any method I know was the identifier. My image is labelled as a II* F Standard converted PNM file. The tiff image now available at the Themis site is headered with this label. II* dz. The gif was labelled gif, the png was labeled png, and the jpeg was labeled jiff.

Might I remind any that PNM is a  well used standard raw satellite data transferral format?

Though by admission I am no expert on these particulars I find it strange that no other image from the Themis site comes with that same tag.

There have been many to attempt to "debunk" this image. No one has, and I have a good hunch no one ever will.

It is plain faulty analysis to take an image and subject it to overtly data torturous nonsense processes designed only to fake a claim that it's a fake based on the observations of said overprocessing steps. The only true way to analyze this image is to process all the data sets and compare. All else is foolhardy supposition using an array of dubious practices for ascertainment. Nonsense in other words

Through experience with processing many more of the Themis images now available it is all too clear that the whole thing, both images and all the testimonies and stories of everyone involved in both studies is pretty much impugnable.

Both the image I got and the IOD are inappropriate for proper processing!

Where did I get this image? From ASU's Themis site, as far as I know. No speculation needed. What did I do to this image? Nothing in any way image destructive. I will stand and defend my stance on this.

Perhaps I was "bamfed" to where a different picture was kept? Perhaps somebody thought it would be a nice joke to play.

I do know this, somehow I obtained a very different tiff image with a dissimilar header identifier which produces superior IR multispectrals over and beyond the presently displayed and original July 24th image of the day. Having looked at some of the example images and visible overlays here on my site you might come to the same conclusion. Personally, I think they are fantastic, opening up an entire new era in Mars exploration.

 My total conclusion is this: The image needs to be released in original unprocessed format and as soon as possible so that more accurate processing therefore deductions can be made. This will continue as soon as it does....

Back to Mars Infrared Imaging