Polar glitches in the MDIM2.1 files?

[Planetary Researcher]

I have been using the MDIM 2.1 files on the web to make map-projected quadrangles. I have noted a couple of glitches, however for the polar regions:

1. The jpg version of MI 70s180E contains errors and will not open in viewers, and from the image dimensions is 128 pixels per degree, not 256 as advertised.

2. The version of the MI83S180E that is on the web is 64 pixels per degree, not the purported 256.

[thare]

for question 1:

We have investigated and are confident that these files are not corrupt. Instead, there are some problems involving the large file size in some software packages. In particular, we are quite sure that there are problems opening large image files of any shape in several web browsers. The results obtained depend on the browser. For example, various versions of Internet Explorer will open the files (slowly) but may not allow one to scroll all the way to the bottom or right. Netscape 7 (Mac and PC) refuses to open the files and gives a very unhelpful error message to the effect that it "cannot open the file because of errors". This result is reproducible with any sufficiently large file, even ones that are known to be OK. It basically means that opening large images in a browser is unlikely to work, but since there is little you can do in the browser anyway (not even see the whole image at once) we do not see this as a strong objection to the size of our MDIM files. We recommend that you download the images rather than opening them in the browser (right-click the link on a PC or hold down the mouse button on a Mac to get a contextual menu that allows you to do this).

The question of what happens once the image is downloaded is more complicated. We have been successful in opening the file you mention in some software packages but not in others. In particular, Adobe Photoshop 7 will not open the file, which is expected because it has a limit of no more than 36,000 lines or samples (the limit on total size is 2 GB, so the 117 million pixels is not a problem, only the image width). This limit should be raised to 300,000 lines or samples in Photoshop CS (aka 8 ) but we get an error message opening the file in this version as well. We are working to try to identify the problem. Strangely enough, it can open a large PNG, so the Jpeg reader seems to have a bug.

In the meantime, irfanview http://www.irfanview.com is a free viewer for Windows that can be used to open the file and crop out pieces that can be opened in Photoshop. Looks like JView will work on the Macintosh. Gimp should work on Windows, Mac, linux and other support machine types. The Windows version of Gimp was able to open these large files.

Other results: Jasc Paint Shop Pro can open and edit the file but is not free. Windows Picture and Fax Viewer can open and display it but not crop it. Microsoft Photo Editor and Apple Preview cannot open it but wouldn't be able to crop it anyway (the last 3 are standard with their respective operating systems). ImageJ is a nice, free, multiplatform program with image editing capabilities, but it could not open the file.

Given that the problems with these files once downloaded result only from the number of samples, not the total image size, and that there exist programs that can work with these files (and there may be more in the future, i.e., if Photoshop CS can be made to do what it is supposed to do), we do not intend to redesign the division of the MDIM 2.1 into files.

MORE:
http://astrogeology.usgs.gov/Projects/MDIM21/

Footnote 4 (on all data distribution pages)
http://astrogeology.usgs.gov/mdim-bin/dataListPage.pl?lat=90N&lon=000E

(posted by pgismgr, written by R. Kirk)

[thare]

for question 1 and 2:

We have adopted a new map projection, equirectangular, for MDIM 2.1. This is described on the main page http://astrogeolgy.usgs.gov/Projects/MDIM21/ but I will try to describe it in another way here. The resolution/scale of an Equirectangular projection is specified at the "center latitude" and the linear dimensions (not degree dimensions) of a pixel are equal at that latitude. Many of our files have center latitude 0, which makes them equivalent to Simple Cylindrical, with equal numbers of pixels per degree of latitude and longitude at the equator. The high-latitude files you commented on, however, have center latitudes chosen to make the map resolution 256 pixels per degree of latitude by 128 per degree of longitude (for the 70N and S files) and 256 per degree of latitude by 64 per degree of longitude (for the 83N and S files). Note that these files are formally defined as having a map resolution of 256 (i.e., the resolution parameter refers to the pixels per degree in the latitude direction). The "256 x 128" description is informal but gives a clearer idea of how the Equirectangular projection actually works.

This approach to projecting the polar data has several advantages, and is being adopted by teams on the Cassini and MRO missions as well as for MDIM:

1) It reduces the file size; our 70N/S and 83N/S files spanning the full 360 degrees of longitude are similar in size to the lower latitude files and to the 25x60 degree Simple Cylindrical files you suggest. Granted, the problems with file width as opposed to total file size were unforeseen.

2) It reduces the total volume of files needed to cover the planet. By having 2 polar files with nonzero center latitudes we can fit the whole MDIM onto a single DVD volume. MRO CRISM will change center latitudes every 5 degrees of latitude and will save many, many DVDs worth of space for their dataset.

3) It reduces the distortion inherent in the Simple Cylindrical projection toward the pole. For example, the 70S file has no distortion at its northern edge and only about 2:1 distortion at its southern edge.

4) Unlike Sinusoidal projection, the Equirectangular files can be converted back to Simple Cylindrical in almost any image processing software, just by resampling to increase the image width (e.g., doubling the number of samples across the 70N/S files while leaving the number of lines unchanged). This is useful if you want to mosaic together data from adjacent files such as 70S180E and 48S090E. If you are not combining the polar data with data from adjacent quadrangles, then you may be better off working with the Polar Stereographic file 90S000E, which is much less distorted.

(posted by pgismgr, written by R. Kirk)