The MGS Face Image in a Different Light

(Revised 8/17/00)

UPDATE 1/26/02 -- Much has transpired since Kelly released his first enhancement of the 1998 MGS image of the Mars Face. As most readers probably know, a new image was acquired in April 2001.  Kelly has not only redone his 1998 enhanceemnt in a more rigorous manner but has applied the same technique to the new image as well. The update is at the bottom of this page, which is otherwise unchanged.

Tom Van Flandern has posted a rectified version of the MGS "Face" image (see Figure 4 on that web page) taken by the Mars Global Surveyor spacecraft in April, 1998. This version, shown in Figure 1, was constructedby Mark Kelly, a computer graphics specialist. It shows the Face illuminated from the upper left, a normal lighting direction for viewing a face as opposed to the totally abnormal illumination in the actual MGS image: the "flashlight-under-the-chin" illumination. If Kelly's reconstruction is accurate, it seems there can be little doubt that this landform is an intentional representation of a face, although perhaps a humanoid rather than strictly human face.But how accurate is it? This article addresses that question with a comparison of his reconstruction with the two Viking images taken by the Viking Mars orbiter in 1976.

Figure 1. MGS Face image processed by Mark Kelly to simulate an overhead view and a lighting angle similar to that of the Viking images.

The original MGS image was taken at an angle 45 degrees off the spacecraft nadir (the line connecting the spacecraft to the center of the planet), rather than from almost directly overhead as in the two previous Viking images. In constrast, Viking Frame 35A72 was taken with a camera line of sight only 10 degrees off nadir and Frame 70A13 had a line of sight 12 degrees off nadir. A negative version of the MGS image retaining the actual perspective of the camera is shown in Figure 2. Note the pronounced elliptical shape of the crater at the lower left. This crater is almost perfectly circular in the Viking images of the same area due to the overhead view of the Viking spacecraft.

Figire 2. Negative of the MGS Face image retaining the original perspective of the camera. Reduced to 26% of the full size.

Part of Kelly's processing was to "orthorectify" the MGS image to produce an overhead view similar to those in the Viking images. His orthorectification procedure involved matching certain points on the original MGS image to the same points seen on the Viking images and then reshaping the image This reshaping requires image processing software that few people have, including myself.While I cannot replicate Kelly's procedure, I can compare his results to the Viking images. Kelly's orthorectified image is shown flanked by the two Viking images in Figure 3. While there is necessarily some lack of precision in matching points in the MGS image with those on the Viking images due to the much lower resolution of the Viking cameras, Kelly's rectification procedure seems to have faithfully reproduced the general shape of the "mesa." The regularity of the landform's perimeter suggested by the Viking images was not an illusion; it persists in the far-higher resolution MGS image.

Figure 3.  Left: Face from Viking Frame 35A72 ; Center: from Mark Kelly's enhancement of the MGS image; Right: from Viking Frame 70A13

Revision to my original evaluation: My most serious criticism of Kelly's enhancement, below, is withdrawn based on a clarification by Dr. Van Flandern:

More problematic is the procedure by which Kelly changed the lighting of the Face from what it really was in the MGS image (sun below and slightly to the left of the Face's long axis) to a lighting intended to be more similar to that of the Viking images (sunlight coming from the upper left). Areas on the right side of the Face close to the "nose" ridge were either extremely foreshortened or hidden altogether due to the off-nadir viewing angle of the MGS camera. Van Flandern states that Kelly "artistically" filled in some of the shading in those regions on the assumption that features on the right would be symmetrical with the features on the left. This has understandably raised red flags in the view of some people, because it is not scientifically permissable to support a theory by making assumptions that derive from that theory. And the only reason to expect such symmetry between the left and right sides of the Face is, of course, based on the hypothesis that the Face is an artificial representation of a humanoid face, which necessarily would have such symmetry.

Van Flandern's clarification, which he has posted to his web site:

"I have since learned that the last two sentences [regarding artistic filling in of features] of this paragraph (immediately following Figure 4) are incorrect. Mark Kelly filled in the small, missing information using low-resolution Viking imagery, not artistically. This was based on Mark Carlotto's shape-from-shading models that combine both Viking and MGS images. Kelly used Figure 2 at Carlotto's site, <>. This answers concerns expressed about the reality of the extension of the mouth to the right (east) side of the Face, and the presence of a complete eye socket on that side. Both are real features, not artist's concepts."

Figure 4 delineates the correspondences of 16 features in Kelly's enhancement with features in Viking Frame 70A13, the Viking image with a sun angle high enough to see some of the features on the eastern side of the Face. Features delineated in green are those that seem, in my opinion, to match well between the two images. Features delineated in yellow are those that seem to be noticeably off. Note that of these sixteen features, there are only three delineated in yellow. In the context of the description of these features, "top" means higher in altitude above the surface of the planet and "bottom" means those parts of the Face at the surface of the surrounding plains of Cydonia while "upper" means features toward the "crown" of the head and "lower" means features closer to the "chin."

The extension of the dark shadow of the "mouth" from the left side of the Face to the right side (Feature #15 in Figure 4) is minimal and Kelly did not impose perfect symmetry on the extension. Viking Frame 70A13 does suggest such an extension, although it appears to angle sharply up from the horizontal trend it has on the more visible left side. This, however, might be the shadow cast by the "nose" ridge on the right side of the Face rather than an extension of the "mouth." Even in the unrectified MGS image  the "mouth" does appear to extend slightly to the right of the "nose." Both Viking images also suggest the presence of an "eye socket" on the right side of the Face, although again, Frame 70A13 suggests the "brow ridge" (Feature #9 in Figure 4) angles up more sharply from the point near the "nose" ridge toward the right edge of the landform. The shape, position,  and orientation of the inner depression of the eastern "eye socket" (Feature #11) in Kelly's enhancement is quite consistent with the Viking image.


Figure 4.  Comparison of  positions and orientations of specific features in Kelly's enhancement of the MGS Face image (left) with the Face in Viking Frame 70A13 (right) as enumerated in the list below.


  1. top western edge of Face platform.
  2. bottom western edge of Face platform/
  3. western side of lower "lip."
  4. western side of upper "lip" and "mouth" cavity.
  5. "teardrop" outcropping.
  6. "Iris" feature
  7. northern apex of western "eye" depression
  8. inner corner of western "eye"
  9. eastern brow boundary
  10. intersection of "headress stripe" with bottom of "headress"
  11. depression or shadowed area in eastern "eye socket"
  12. top eastern edge of Face platform
  13. edge of "terrace" on southeastern part of the Face platform
  14. bottom eastern edge of Face platform
  15. eastern side of "mouth"cavity
  16. tip of "nose" (more like a beak in the lower-resolution Viking image)
As I learned on 8/17/00, the features in the Kelly enhancement that do not match those in Viking Frame 70A13 are not the result of any "artistic license" on the part of Mark Kelly. Instead, the features I delineated in yellow might be regarded as "second order" features -- those visible in Carlotto's enhancement of the Viking images based on the MGS data, but not readily apparent in any of the original Viking or MGS images. The features delineated in green could then be considered "first-order" features that can be directly seen to match between the MGS and Viking image. The validity of the second-order features may be more uncertain than that of the first order features, but they are not the result of any arbitrary decisions on Kelly's part. Figure 5 below shows the Kelly enhancement next to the image Carlotto derived from his shape-from-shading model for late afternoon sunlight. Feature #8, the inner corner of the western "eye socket," which seemed not to match the position in the Viking image, has been omitted in Figure 5, because its relevance to the question of facial resemblance and symmetry seems fairly trivial. The eastern "brow" line (feature #9) in Kelly's enhancement appears to match Carlotto's well. There is a suggestion of an extension of the mouth on the eastern side of the Face (feature #15) in Carlotto's image, although this seems more indistinct in my opinion.The slant of the eastern "brow" ridge, which appears noticeably asymmetric with the western "brow" ridge in Viking Frame 70A13, may simply be the result of the direction of the sun in that particular image, perhaps combined with some degree of real asymmetry in the landform. Some of the finer details in Kelly's enhancement are not visible in the Carlotto image strictly because the Carlotto image is primarily derived from the lower-resolution Viking image; these features are all visible in the original higher-resolution MGS image.

Figure 5. Comparison of the Kelly enhancemt  (left) with an image Mark Carlotto generated from his shape-from-shaind models for the Viking images refined using the MGS image.

While there is still some uncertainty in Kelly's reconstruction, he has clearly not tried to hide the apparent asymmetry in the areas delineated in yellow in Figure 4. What he shows suggests an artificial object that has been damaged on its eastern side -- a possibility suggested long ago by Mark Carlotto and other proponents of the artificiality hypothesis based on the second Viking image. Given the uncertainties caused by the low resolution and shadowing of the right side of the Face in the Viking images, Kelly's reconstruction still may prove accurate if NASA ever fuflfills its promises to acquire additional images of the Face.

As for the features on the left side of the Face, which were much clearer in the Viking images, the shadowing of the "eye socket" and "mouth" in Kelly's reconstruction are very consistent with the Viking images. The larger shadow of the left eye socket is more consistent with Viking Frame 35A72 than with the higher-sun elevation image, Frame 70A13. For the most part, he has shown not what the Face "might" look like but what it mustt look like in an image made with the lighting and viewing geometry of the Viking images but with the 10-times higher resolution of the MGS camera.

Comparison of Figure 1, the Kelly enhancement, with Figure 2, a negative of the orignal MGS image, shows that Kelly did not remove any of the rough, "fractal" features that could be seen as support ing the interpretation of the Face as a natural formation. They are present "warts and all." With the possible exception of the extension of the "mouth" on the eastern side of the landform, no features have been added that were not present in the original MGS image. The strong resemblance to a face in this enhancement cannot, therefore, be attributed to any digital trickery such as "airbrushing" out unwanted natural features or "morphing" the image with an image of another face.

And it should be emphasized that the nostril-like depressions at the end of the "nose" ridge, while not visible in either of the low resolution Viking images, are undeniably real features in the MGS image. They were not added by Kelly. It is the presence of these "nostrils," perhaps more than anything else that gives a strong impression that Kelly's reconstruction shows an artificial object. In The Case For the Face, published before the MGS image was taken, archaeologist Dr. James Strange suggested that nostrils would be among features that the MGS might see if the artificiality hypothesis were valid. Kelly's reconstruction graphically demonstrates the success of this prediction, which would be highly unlikely if the Face is the result only of natural weathering and erosion. Kelly's work may not be the final answer, but I think it is a valuable contribution to Cydonia research -- perhaps the most complete synthesis of the Viking and MGS data possible at present.

UPDATE 1/26/02

A few months before the release of the April 2001 MGS image of the Mars Face, Mark Kelly made a second  enhancement of the 1998 image with the same purpose as the first: to show what the Face would look like at the high resolution of the MGS camera but under the lighting conditions of the Viking images, which were closer to the lighting under which faces are normally observed -- from above the top of the head rather than below the "chin" as was the case in the actual 1998 image.

In this second enhancement, he did not draw in the shadowing of the Viking image by hand as he did for his initial version. Instead, he used Photoshop to overlay a semitransparency of the Viking image directly onto the an orthorectified version of the MGS image. The overlay technique is a mechanical process that greatly reduces the freedom for subjective judgements that is inevitable when drawing by hand.

Also, the orthorectified version of the MGS image is more accurate than the one he used for his first version. The problems I described previously with his first enhancement were primarily associated with spatial positions rather than image contrast and brightness.  The most significant problems were the extension of the "mouth" to the east side and the overly-precise alignment of a possible "eye" socket on the eastern side relative to vertical position of its western counterpart, features that I felt were not supported by the actual image data. These problems were corrected in Kelly's second enhancement.

For a time, Kelly had this enhancement on his web site, Shortly after he posted it, the 2001 image was released by NASA, so Kelly's improved enhancement went unnoticed by most people.  Since then, he has replaced the enhancement of the 1998 image with a new enhancement of the 2001 image using the same overlay procedure.  For comparison of the two enhancements, they are shwon side-by-side versions of Kelly's enhancements below.

Mark Kelly's Enhancements of the Mars Face. The enhancement of the 1998 image is on the left. The enhancement of the 2001 image is on the right.

The orthorectification of the 2001 image was done by Malin Space Science Systems, not by Mark Kelly himself. The viewing angle was much closer to an overhead view to begin with for the 2001 image (25 degrees off-nadir versus 45 degrees in the 1998 image), so there is (or at least should be) much less uncertainty about the shape in the 2001 image. Somewhat ironically, the "nose ridge" is markedly more symmetrically positioned in the MSSS orthorectification of the 2001 image than it was in Kelly's ortho of the 1998 image, where it is noticeably to the (viewers) right of the vertical centerline. So in this regard, MSSS gives more support for an artificial interpretation of the landform than does Kelly, albeit inadvertently.

While the 2001 image provides a more reliable depiction of the landform's shape as it would appear when viewed from directly overhead, the 1998 image is probably more reliable when it comes to image contrast and tonal variations. The sun's elevation in the 2001 image was 52 degrees, an angle so high that the image has a washed-out appearance. Note that not only is the eastern "nostril" feature clearly visible in the 1998 image missing in the 2001 image, but also some of the other bumps and ridges visible in the 1998 image are gone as well.  The disappearance of these features is very likely due to the high sun angle in the 2001 image.

Another difference between the 1998 and 2001 enhancements is that Kelly did the VIking overaly on a negative of the 1998 image while he used a positive of the 2001 image. The use of a negative was justified by the fact that sunlight was coming from very nearly the opposite direction in the VIking image, and dark and light surfaces should, in fact, be reversed for the VIking illumination conditions. A weakness of using the negative, although probably a minor one, is that dark shadows are simply made light-colored rather than being reversed in direction as they would be in reality. Since there is so little tonal variation the 2001 image, it provides a more neutral surface on which to superimpose the tonal variations of the Viking image.

The way the landform _really_ looks under the lighting conditions of the Viking images at the resolution of the MGS camera is probably somewhere between the two enhancements.

 -- Lan Fleming


Back to VGL Mars Page

Related Articles:

JPL's "Catbox" Enhancement of the "Face"

A comparison of the Face to Terrestrial structures.

Evaluation of the "Eye" and "Mouth" Features of the Face Mesa

VGL Home Page