The field of digital spacecraft imaging and image enhancement is quite unlike ordinary photography and photographic processing, and is generally unfamiliar to the public, where the tendency may be to think of the spacecraft images as similar to photographs taken with an ordinary camera.
This unfamiliarity has at times been capitalized on by critics of the Mars anomaly research to selectively cast doubt on only those digital image enhancement processes that have provided evidence that certain Martian objects, digitally imaged by NASA'S Viking mission, may be artificial. (It should be noted here that Mars anomaly researchers have not claimed these items of evidence constitute "proof" of artificiality, but only that they increase the probability, making further investigation necessary.)
After the article "Dr. Malin's False 'Teeth'" was published recently in this newsletter, the following criticism was forwarded to me from the internet. The critic writes that he cannot "see" the "teeth" found in Dr. Mark Carlotto's enhancements (not the same as the "false teeth" identified by Dr. Malin) when he looks in the "raw data." He then argues that this *proves* the "teeth" are merely artifacts of digital enhancement and not real features.
This objection raises the issue of the relationship between "raw data" and "enhanced" images.
Actually what the critic referred to as "raw data" is not "raw" at all but has been improved for viewing by increasing its contrast (the raw Viking image has very low contrast and its contents are hard to discern otherwise). Additionally, the critic employed "pixel replication" to enlarge the image; however, this form of enlargement has no effect on the resolution of the image. It merely makes it larger, and therefore easier to view. A pixel is the smallest unit of information in the digital image, whose brightness value is transmitted as a binary number and represented visually as a small square.
Figure 1. Simple pixel replication. The new pixel is the same shade as the original.
Although technically not quite correct, it is reasonable to refer to the image after pixel replication and contrast stretch as "raw data" because little is changed in the image other than to enlarge it and apply appropriate contrast control. Hereafter when we refer to "raw data" (in quotes) we will mean the image after pixel replication and contrast control. The image prior to these modifications will be referred to as raw data (without quotes).
There are examples of both raw data and "raw data" (the result of pixel replication and contrast stretch) in Unusual Mars Surface Features by DiPietro, Molenaar, and Brandenburg.[1] Their Figure 3 (page 16) shows an image of Viking frame 35A72 identified as raw data. Obviously in this image the contrast is very bad and only the gross features can be made out. In their Figure 7 (page 21) they provide a contrast-stretched and digitally enlarged (by pixel replication) image of 35A72. This is evidently the same as what the critic did to get his "raw data" in which he does not "see" the teeth.
However, contrary to his claim, in this image there is clearly a fairly complex set of variations in pixel values in the mouth area, covering about 40 pixels. [2]
Figure 2. Complex variations of pixels in the mouth area of the Face. Contrast stretched and enlarged by pixel replication. Bright pixels in dark area are transmission errors.
One would certainly not expect to see the "teeth" in the "raw data" exactly as they appear in enhanced frames. Thus the critic is setting up a condition impossible to meet, namely that he should see the "teeth" exactly as they appear in enhanced frames, whenever he looks at an unenhanced frame. This objection is more of a rhetorical "trick" than a meaningful observation. What we do see is these pixel variations covering a relatively large area (Figure 2). There is something there, possibly a complex structure.
The scientific question to ask is, "what does this pattern of pixel values in the "mouth" yield when effective pixel interpolation algorithms are applied?"
Each pixel, represented visually as a square area containing a single grey-scale value, derives from
a binary number representing the integrated average shade, or "value," of all the details in
the area it covers.[3]
Figure 3. In this figure the "recorded data" represents a portion of one single line that the camera records as it passes over the object. For emphasis colors have been used to represent grey scale values. (Taken from Unusual Mars Surface Features, page 23.)
As the diagram shows, an integration of values occurs where the edges of the darker object fall within the area of the camera's pixel definition. In the figure, the camera records a value of 66 instead of either 88 or 44. In the recorded ("raw") data the sharp edges of the object are "lost" and the object may be unrecognizable. The complex set of pixel shades shown in Figure 2 above represents such an averaging, or integration, of values as the camera responds to variations in the actual contour and shadows of the Face's "mouth" area.
However, the relations of values between the pixels contain more information about the original shape of the object than is visible to the eye. Because the specific variations of pixel values were originally derived from an averaging of the values in the actual scene, with an appropriate interpolation algorithm the original shape of the object can often be recovered with reasonable accuracy. According to DiPietro and Molenaar, interpolation actually amounts to a statistical analysis of the data provided by the pixel relationships.
Such a procedure involves the original pixel and the values of immediately
surrounding pixels. The original "large" pixel is divided into smaller pixels, whose new
values are derived (interpolated) from the partially-shared values of the adjacent original
pixels.[4]
Figure 4. Bilinear Intepolation. The replacement for the original pixel takes its values from surrounding pixels. NOTE: Colors have been used to represent the grey values for emphasis.
Bilinear interpolation utilizes only the four adjacent pixels, while more sophisticated algorithms may reference more. Cubic Spline interpolation, the one used by Dr. Carlotto, utilizes weighted values from twelve surrounding pixels. The Starburst Pixel Interleave Technique (SPIT) utilized by DiPietro and Molenaar uses data from eight surrounding pixels.
The interpolation algorithm examines each pixel of the raw data one by one and queries all of the surrounding pixels, which cast their weighted votes to produce values for new pixels to replace the one raw data pixel. The process then moves on to the next raw data pixel, and queries the adjacent pixels around it to redefine new values for this raw pixel, and so on.
If the interpolation algorithm employed is one that has been tested for accuracy against actual features, and it is used with appropriate care, the result is to yield information inherent in the relationships between the various pixels, which is ultimately a function of the way the camera data is gathered in the first place.[5]
Figure 5. The processed data yields an image closer to the original than the "raw data.".
Thus, for example, the genuine circularity of a crater rim can be recovered when an appropriate interpolation algorithm is applied, even though in the "raw data" the shape is visually unclear. A dramatic example of this result is shown in Unusual Mars Surface Features (DiPietro et. al. 1982, 1988) on pages 34-35. A "raw data" Landsat image of the Pentagon building is placed side-by-side with an enhanced version. The cell-like structure and pentagonal shape of the building, indiscernible in the "raw data," is clearly visible in the enhanced version, which was processed using the "Starburst Pixel" method of interpolation. Indeed, looking at the "raw data" alone one sees something similar to what is shown in Figure 2 above: There is something more complex there, but without enhancement its shape is unclear.
Production of artifacts (false data) is always possible if such algorithms are improperly applied, but this does not mean that enhancement by interpolation is a useless procedure or that under controlled circumstances it cannot bring out real features that are not clearly visible in the "raw data." On the contrary, interpolation in competent hands is in regular use for clarifying digital imagery.
In view of this, turning to the "raw data" alone is a step backward in the scientific procedure of analyzing digitally encoded spacecraft images.
When a sophisticated interpolation (such as "cubic spline") is applied to the Face on Mars, what we discover is a set of features in the "mouth" that resemble "teeth" (Figure 6). There is no reason to suppose that this feature is spurious, particularly since it appears with the same contours in both Viking frames and in response to more than one method of interpolation.
Note: This image is for reference purposes only. Quality may vary depending on your hardware/software. For high quality reproduction see The Martian Enigmas by Dr. Mark J. Carlotto. Note also that the "teeth" shown here are not the same feature as that pointed to in error by Dr. Michael C. Malin. The latter are overenhanced individual pixels in a different location. See "Dr. Malin's False 'Teeth'" in this newsletter.
As mentioned earlier, the "teeth" feature appears in both of the available higher resolution images (Viking 35A72, 70A13) and throughout several methods of interpolation -- a fact that makes it extremely difficult to deny the reality of the feature on the grounds that some processing error has occurred.[6] But when this supporting evidence for the reality of the feature was pointed out to the critic, his response was to argue that some processing error must have occurred prior to interpolation and thus affected all subsequent processing.
Since there is no particular evidence to support this claim other than the apparent determination to believe that the "teeth" are unreal, this kind of objection involves the logical fallacy of a priori reasoning -- assuming in advance that the "teeth" feature can't be there, so evidence that it is a real feature is simply rejected.
Unfortunately the claim that an error "must have been" introduced prior to interpolation is not consistent with the critic's claim that it is the interpolation that produces the alleged errors. Nor does it deal adequately with the fact that the same features appear in both frames, which are differently oriented to the pixel grid -- precluding the same error occurring regardless of what stage of enhancement was involved. Nor is any suggestion provided as to how, and at what stage, this could happen. What we seem to see in this approach is a need to insist the feature is spurious at the expense of rational inquiry.
In this we also appear to have a parallel with the behavior of a planetary scientist as described in The McDaniel Report (page 169): When Dr. Carl Sagan was being shown the first enhanced photographs of 70A13, another planetary scientist present at the meeting refused to look at the images, actually placing his hands over his eyes in order to avoid seeing them -- much in the same way the Churchmen refused to look through Galileo's telescope.[7] Anthropologist Randolfo Pozos (Author of The Face on Mars) states of this type of reaction:
"It was almost as if people immediately assessed the potential damage to their belief systemàand established immediate defensesà"
Another symptom, perhaps, of this element of denial is a further objection that was initially made by Dr. Sagan in his infamous "Man in the Moon" article in Parade Magazine (1985) and has been echoed frequently since. Dr. Sagan implied that Mars anomaly researchers were peering at images "at the very limit of resolution" and then trying to make sense out of vague or indistinct data. This objection ignores the fact that the major features of interest on the Face significantly exceed the limit of resolution. The limit of resolution, of course, is one pixel, or about 2000 square meters for Viking frames 35A72 and 70A13. In comparison the "mouth" area where the "teeth" are visible is at least 80,000 square meters -- 40 times larger than "the limit of resolution."
Of course, this does not mean that we can obtain a perfectly clear and detailed image of the "teeth." It is important to keep in mind that none of the legitimate Mars anomaly researchers has claimed the feature is known without any doubt to represent teeth. However enhancement reveals an apparently complex structure that bears sufficient resemblance to "teeth" to raise the question of artificiality -- especially when seen in the context of the other significant features of the Face and nearby anomalous objects.
The potential significance of what can be seen in the "mouth" area seems to be well-testified to by critics, when one considers the effort put out in attempting to discredit the feature.
This phenomenon of denial and often vehement psychological resistance to any data that may upset an individual's world view was noted in the Brookings Institution Report mentioned in the current excerpt from The McDaniel Report in "Watch This Page..." However, adamant and unreasonable skepticism regarding Mars anomaly research is difficult to understand when it is clear that the legitimate researchers have never claimed that the data constitutes "proof" of artificiality. The claim is merely that there is sufficient data to warrant giving high priority to further investigation. Unfortunately it appears that there are some whose aversion to the possibility of artificial objects on another planet is so great that they oppose even taking a closer look.
FOOTNOTES
1. For this and other references see "Where to Get More Information."
2. This same complexity is also visible in the top image on page 15 of Dr. Mark J. Carlotto's book The Martian Enigmas. In that image, where again only pixel replication is employed, the resemblance of the complexity in the mouth area to the "teeth" visible after further enhancement is fairly noticeable. Note that since the limit of resolution is one pixel, the structure is not "at the limit of resolution" as some claim.
3. See Figure 4, page 17, in Unusual Mars Surface Features by DiPietro, Molenaar, and Brandenburg. Each pixel is derived from an eight-digit binary number representing a shade of gray from total black to total white in 256 gradations.
4. See the bibliography below for supporting references.
5. An explanation of this is given in Unusual Mars Surface Features, pages 16-34.
6. See Carlotto, Mark J., The Martian Enigmas, pages 14-15.
7. This amazing incident was reported to me by Dr. David Webb, former member of the National Commission on Space under President Reagan, who was a witness to the event.
BIBLIOGRAPHY
V. Tom, R. Merenyi, M. Carlotto, and W. Heller, "Advanced enhancement techniques for digitized images," 15th Symposium on Nondestructive Evaluation, April 1985, San Antonio, TX.
W. Schreiber, "Image processing for quality improvement," Proc. IEEE, Vol. 66, No. 12, Dec. 1978.
T. Peli and J. Lim, "Adaptive filtering for image enhancement," Optical Eng., Vol. 12, No. 1, 1982.
D. Wang, A. Vagnucci and C. Li, "Digital image enhancement: A survey," Computer Vision, Graphics, and Image Processing, Vol. 24, pp 363-381, 1983.
V. Tom, "Adaptive filter techniques for digital image enhancement, Proc. SPIE, Vol. 528, 1985.