A preliminary evaluation is presented of the April 5 image of the Face in the Cydonia region of Mars acquired by the Mars Global Surveyor (MGS). It describes different cleanup, enhancement, and orthorectification processes than those used by JPL. The resulting image contains more tonal diversity making it easier to analyze topographical features. By using elevation information derived from Viking frame 70A13, we have also been able to obtain a better approximation to an overhead view than that provided by JPL.
High resolution MGS imagery over Cydonia was posted on the Internet at 1:30 PM Eastern Daylight Time, April 6. The raw image (cydonia1.tif) is a strip 1024 pixels wide by 9600 pixels long. Figure 1 shows the portion of the high resolution strip containing the Face after being flipped left-right and contrast stretched.
Figure 1 Raw image flipped left-right and contrast stretched. (NASA)
Click here for full resolution version
The raw image contains vertical striping caused by sensitivity variations along the horizontal direction in the photodetector array. The striping pattern was removed by computing the average pixel brightness of each column in the strip, dividing the brightness of each pixel in the column by the average brightness of the column, and multiplying the result by 128. The resultant destriped image is shown in Figure 2.
Figure 2 Destriped image. (NASA/Carlotto)
Click here for full resolution version
On close inspection the image contains several bright areas which appear to be partially obscured by thin clouds or haze. Also evident are variations in surface brightness. Both of these effects confound the estimation of local topography from shading information. One solution is to apply a high pass filter to the image. However if the low-frequency cutoff is too high (the spatial extent of the filter is too small), important tonal information will be lost making it difficult to understand the shape of objects. Figure 3 is a high-pass filtered image computed by subtracting a Gaussian smoothed version of the image from itself. The standard deviation of the Gaussian (sigma=64) was chosen to remove most of the background variations while preserving local shading information.
Figure 3 High-pass filtered image. (NASA/Carlotto)
Click here for full resolution version
The 1976 Viking images provide a visual basis for analyzing this new image of the Face. Since the Viking and MGS images were acquired under different lighting conditions and from different perspectives it is necessary to optically transform and geometrically reproject the data in order to compare the two images.
To do this it is necessary to know the new MGS image acquisition parameters. They are as follows:
|
Solar Elevation |
25 degrees |
|
Solar Azimuth |
159 degrees |
|
MGS Elevation |
45 degrees |
|
MGS Azimuth |
232 degrees |
Figure 4 shows a cleaned up and enhanced version of the MGS image. Next to it is a computer generated rendering or prediction of what MGS would have seen based on the shape of the Face derived from Viking frame 70A13. It was produced by first computing a shaded rendition of the Viking 70A13 elevation surface using a Lambertian reflectance function for the MGS sun angles, and then projecting the image to the position in space where MGS was located at the time it acquired the new image of the Face.
Figure 4 MGS image of Face (left) and predicted view generated from Viking frame 70A13 (right). (NASA/Carlotto)
The Face is illuminated from below and so the chin is bright and the mouth is in shadow. The ridge line of the nose is partially illuminated. Although the top of the head is sloped away from the sun, its slope is gradual enough to be illuminated.
Figure 5 MGS image of Face (left) and reprojected image from 70A13 (right). (NASA/Carlotto)
Figure 5 compares the MGS image with the 70A13 image of the Face reprojected to match the MGS acquisition geometry. The bright area near the left eye in Figure 4 corresponds to a surface sloped toward the sun in the MGS image. In the Viking image in Figure 5 the illumination is from almost the opposite direction. The surface thus slopes away from the sun in that view and is dark.
Many points of correspondence can be seen in Figure 5. The middle of the mouth seems to come together in both views. The two bright circular features in the MGS image which look like nostrils correspond to a flat area near the tip of the nose ridge in the Viking image. Subtle brightness variations on the left side of the base of the Face first seen in the Viking images are visible in the MGS image and appear to be variations in surface brightness. This pattern seen as broad stripes in the Viking images extends across the Face and correlates with narrow ridges or lineaments on the left side of the nose ridge. It is not clear if the left eye seen in the Viking image corresponds just to the dark circular feature on the left side of the Face in the MGS image or if perhaps the sloped surface above it also contributes to the visual effect. The crossed lines on the top of the head seen in the Viking imagery are very prominent in the new MGS image. These lines, the symmetry of the base and the precision of its beveled edge are perhaps the most striking features of this formation. Whether it is a face or not remains to be determined.
Following the initial imagery release, JPL released a geometrically stretched version of the original image that simulates what the Face would look like from above. Orthorectification is a process by which an image that is acquired obliquely is reprojected to appear as if it was acquired from directly overhead. Only if the terrain is flat can orthorectification be done by simply stretching the image to compensate for the foreshortening in the direction of the observer. Because the topography of the Face is not flat stretching the image will actually distort the internal structure of the Face, perhaps rendering it less face-like.
The MGS image was taken from the west, 45 degrees off-nadir, and so much of the right side of the Face is obscured. As a result it is not possible to obtain an accurate representation of its shape from the MGS image. Fortunately the Viking images were taken overhead at low sun angles ideal for shape from shading. Figure 6 is the MGS image of the Face, and the elevation map derived from Viking frame 70A13 reprojected to match the Face's perspective as seen from MGS.
Figure 6 MGS image of the Face (left) and the elevation map derived from Viking frame 70A13 (right). (NASA/Carlotto)
Figure 7 shows the stretched version of the Face provided by JPL. Because it is being viewed from the left, the left side of the Face is much wider than it should be while the right side is much narrower. Using Viking's 70A13 height map, a much better approximation to the overhead view can be obtained. In this improved version the nose ridge is centered laterally on the face, and the internal features of the Face appear much more symmetrical. Because there is very little information on the side of the Face pointed away from MGS, there is very little detail on that side as seen in the figure. Although there is a faint indication of an eye feature on the right side of the Face, final judgment must await additional imagery.
Figure 7 Stretched version of the Face provided by JPL (left) and better approximation to orthorectification computed using 70A13 height map (right). (NASA/Carlotto)
In the 1976 Viking images, the impression of a face was unmistakable. But illuminated from below the Face seems less remarkable. Many of the features seen in the Viking images of the Face can also be seen in the new MGS image. But it is apparent that the Face, if it is in fact a face, is severely eroded. Although the symmetry of the platform, precision of the beveled edge, linear features at the top of the head are striking, perhaps they can be explained geologically.
However, before a final judgment can be made, several key questions remain. First, are there facial features on the right side of the Face? If so their presence would greatly increase the probability that this object is a face, regardless of its present condition. Second, how does the Face appear at high resolution under the original Viking imaging geometry with the spacecraft almost directly overhead and the sun to the northwest? Like any face, one's subjective impression of it varies with the illumination. Perhaps the Face was designed to be viewed under more restrictive illumination conditions.
The most important question is, what do the other objects look like at high resolution? For example, finding the straightness and geometrical regularity in the walls of the Fortress to have an obvious geological explanation, or the mounds to be naturally occurring landforms that happen to be arranged in a regular pattern would weaken our hypothesis that this collection of object is artificial. Being more geometrical in shape, these objects, if they are artificial, may have retained enough of their architecture to be recognizable, even in a degraded condition. Hopefully future MGS images will help answer these and other questions about this strange collection of objects on the surface Mars.
Figure 8 Simulated stereo view generated from the MGS image reprojected as it might appear from the southwest (top) with a similar view of the 70A13 data shown for comparison (bottom). (NASA/Carlotto)
(Revised 4/14/98)
