More detailed information on how the Ganymede map was created and its accuracy

This section contains rather detailed information on how I created the map of Ganymede. It is somewhat technical and parts of it may be difficult to understand (possibly incomprehensible !) unless you have some knowledge of the images returned by the Voyager and Galileo spacecraft used to create the map. Before you read further, it should be noted that this map is primarily intended for use in 3D/animation work. However, I do not rule out that it might be scientifically useful as well (especially for diagrams or illustrations). That was not my goal though but I tried to make the map as accurate and realistic as I could. You have been warned ;-).

For creating the map I used 18 Voyager 1 images, 44 Voyager 2 images and 14 Galileo images (2 E6 images, 11 C9 images and 1 E14 image). In addition I used a big mosaic of the Osiris area I got from Constantine Thomas. That mosaic was created from 14 Voyager 2 images. He also gave me various ideas, tips and information for which I am grateful. In addition to this I used 7 lower resolution global images (1 from Galileo and three from each of the Voyagers) to colorize the map. This is a total of 21 image for the color stuff since each color image is composed from three images taken through color filters. All in all this is 111 images (!!); needless to say I lost the count a long time ago and didn't know the total number of images until I created this page and was forced to find out ;-). All of these images are "raw" images from the spacecraft. The resulting map is 7200x3600 pixels (2.3 km/pixel); the version available here is 1800x900 pixels. This map has only small areas with fictional or inaccurate data. When creating the map I used a 4 km/pixel (4137x2070 pixels) map of Ganymede from the US Geological Survey (USGS) as a reference.

Positional and photometric accuracy
The map was created from scratch by reprojecting the individual spacecraft images to simple cylindrical projection, compensating for the varying illumination on the fly. Following this I made several mosaics from the individual "datasets" (Voyager 1, Voyager 2, and Galileo). It was necessary for me to calibrate the Voyager images using flat field and dark frame images taken for calibration purposes. The reason is that the raw, uncalibrated Voyager images have poor photometric accuracy, for example they are very obviously brighter near the corners than near center. Calibrating the Galileo images was not necessary for my purposes. I also made no attempt to correct for possible geometric distortion of the raw images.

I used mainly clear filter narrow angle images. For Voyager 1 I used blue filter images in a very few cases where the corresponding clear filter images were smeared. For Voyager 2 I used clear filter narrow angle images exclusively except for a single orange filter image and a single wide angle view of the southern hemisphere. However, the latter two were very important since they enabled me to get relatively good coverage of a region WSW of Osiris that looks very fuzzy in the USGS Voyager-based map. The wide angle image also covered a large area very near the south pole. I also managed to get a nice looking Voyager 1 coverage of the region near the crater Tros. In the USGS map this area is smeared but I was able to use different images that are "theoretically" of lower resolution than the images used in the USGS map but because of the smear the USGS map is not of very high quality in this region (although positionally it should be very accurate).

For the Galileo images, reprojecting to simple cylindrical projection should have been straightforward since the raw image files contain information on the viewing and lighting geometry. However, there were errors (or very large inaccuracies) in this information in the E6 images and some inaccuracies in some of the C9 images but I managed to sort this out. For the Voyager images things were more complicated since no information on the viewing or lighting geometry is available (or so I think; if this statement is wrong I would like to know !). In this case I used the 4 km/pixel map from the USGS to determine the latitude and longitude of several points in each Voyager photo. This makes it possible to determine the viewing geometry and in a perfect world the accuracy of this procedure would be 100%. Unfortunately, my lat/lon measurements are not perfect and there is probably a minor geometric distortion in the images as well, most pronounced near the corners. The USGS map is probably not positionally perfect either. In addition I used David Seal's Solar System Simulator to determine the (approximate) lighting geometry. Despite the above sources of inaccuracies the positional accuracy of my map compared to the USGS map is approximately 5 pixels or better in lat/lon (at 7200x3600 pixel size) in most cases where this could be determined, except in the longitudinal direction near the poles.

After creating mosaics from the datasets I "manually" adjusted the brightness of individual areas within the mosaics using Photoshop since it was impossible for me to compensate to 100% accuracy for the varying illumination of the different images when reprojecting them. I then merged the maps into one big map and removed the seams between them, again using Photoshop (mainly the rubber stamp and gradients for those of you who know Photoshop).

A problem then emerged: The map was created from relatively high resolution images and in these images global brightness variations are overshadowed by local brightness variations. This meant that Ganymede's bright polar caps were far too faint in my map and in fact the southern cap was mostly absent. I therefore "manually" increased the brightness of the northern cap, especially in northern Galileo Regio. Since the southern cap was absent I had to determine its extent as a function of longitude visually from global images. In doing so I "discovered" that its extent varies approximately sinusoidally with the maximum extent occurring in the leading hemisphere. I then brightended everything south of this "sinusoidal curve" and added a "transition zone" north of it. Most of this processing was too specialized to do easily in Photoshop so I wrote small command line utilities to do some of this specialized processing.

After this I applied gamma correction to the entire map to brighten it since it looked too dark.

The final step was to fix areas for which no photos of decent resolution are available. This applies only to relatively regions very near the poles and to a region in the north near 200 degrees longitude.

As the above description shows it may not be a great exaggeration to say that the map's photometric accuracy is close to zero. Despite this, renderings created using the map are very similar to spacecraft photos, both global images and higher resolution images. And the positional accuracy is good.

After this all I had a big grayscale map and all that remained was to colorize it.

For colorizing the map I used several low resolution color images. I composed these images from grayscale images taken through different color filters. For the Voyager spacecraft I used orange, blue and violet as red, green and blue and for Galileo (C10 images) I used IR-7560, green and violet. I used only a small section of the Galileo image. The reason I used mainly Voyager images instead of Galileo images is that Galileo's color coverage is very "unbalanced". For some areas only IR-9680 is available to use as red. Using it as red results in very inaccurate color and most of these images are also very noisy. In some areas red or IR-7560 (both of which I could use) is available but there are also areas with no color coverage.

A montage of the images I used can be seen here. The red/pink color in the bright crater in the Galileo image at lower left is present because the green and violet images were saturated at this spot. In addition I used this Voyager 2 image.

I then reprojected these images to simple cylindrical projection and merged them into one "color map". I then used this color map to colorize the big grayscale map. One unexpected problem was that it turned out to be extremely difficult to adjust the color in all the images to get a continually varying color (not brightness, needless to say it is variable) across the color map instead of sharp variations where two images overlap. This problem was especially nasty where the Voyager 1 and 2 images overlap. The result is that the color variations in the final map may be inaccurate. Galileo Regio looks reddish compared to an area in the left half of the map which looks slightly greenish. I doubt this is correct. This is even more apparent in the polar caps if they are rendered pole-on. In fact I had to "clone color" in some areas near the poles to make this problem less prominent.

Comparison to the USGS map
As previously mentioned the difference between my map and the USGS map in the position of features is usually about 5 pixels or better except near the poles where the difference in longitude is in many cases more than 5 pixels for obvious reasons. The following table shows several semi-random examples. The comparison applies to a 7200x3600 pixel map, i.e. I resized the USGS map to that size for this comparison.

Feature Coordinate (USGS map) Coordinate (my map)
Crater SSW of Osiris (3784,2719) (3781,2718)
Crater east of Memphis Facula (4710,1470) (4708,1470)
Bright spot in NW Galileo Regio (3913,608) (3913,603)
Crater SSW of Gilgamesh (4548,3300) (4528,3289)
Crater in the far south (2529,3343) (2520,3344)
Crater in the north (215,348) (210,347)
Small bright crater near the equator (557,1969) (557,1968)
Small crater in the southern hemisphere (959,2612) (958,2612)
Small bright spot in Perrine Regio (6543,1168) (6538,1168)

Back to Ganymede page