At the present writing, Microsoft's terraserver-usa.com web site provides free downloads of USGS images in one meter resolution for the United States. TerraServer may be accessed either via an interactive web site visited by users, or as a web service to which applications like Manifold may connect to automatically fetch desired images. See http://terraservice.net or http://www.terraserver-usa.com for details. As always, URLs are subject to change and often will change after documentation like this is written. If the URLs in this topic are no longer valid, use a good search engine to hunt down the new locations.
Important: The very cool Microsoft TerraServer should not be confused with the various non-Microsoft enterprises (such as terraserver.com) that have nothing to do Microsoft even though they have managed to obtain confusingly similar domain names.
TerraServer may be accessed either via an interactive web site visited by users, or as a web service to which applications like Manifold may connect to automatically fetch desired images.
§ This topic shows how to work with images obtained using the interactive TerraServer website and how to manually assemble them into larger mosaics.
§ The Linked Images from TerraServer topic describes how to use the built-in Manifold capability of automatically downloading and combining images from TerraServer.
For most purposes, using the built-in Manifold ability to fetch and assemble TerraServer images automatically is much more convenient than manually visiting the web site and then downloading and assembling images by hand.
Nonetheless, we have provided this example using TerraServer because getting images from TerraServer is similar to how they may be obtained from other sources. It also provides a set of images that are easily georegistered using the Assign Projection dialog so that the topic can teach the simple cut-and-paste methodology of making mosaics from georegistered images instead of getting bogged down into lots of discussion on how to georegister images.
The downloadable images provided by Terraserver are smaller than the original USGS DOQ images from which they are generated. To create an image that covers a larger area one must download several adjacent images from Terraserver and then mosaic the images together into a single, larger image.
This topic shows how to download two adjacent images and to mosaic them together. Although the specific steps illustrated are set forth for Terraserver, the general ideas and methods can be applied to piecing together images from other public access web sites from which images may be obtained.
As a general comment, there is not much to creating image mosaics in Manifold if the images involved are correctly georegistered. Simply copy and paste images together to make a larger image. Most time in such projects is usually spent finding the images, downloading them and georegistering them. The actual assembly of multiple images into a single larger image is trivial.
Step 1: Download the images
Microsoft's TerraServer site allows users to download the images displayed in .jpg format. It also allows download of a "world file" for each image. The world file provides partial projection information for the image.
Visit http://terraserver-usa.com/ and locate the area of interest (this URL was active at the time this topic was written). For this example we will zoom into Palo Alto, California and use a USGS aerial photograph taken on October 30,1991. We zoom into the photograph to four-meter resolution for our area of interest. A screen shot showing part of the Terraserver web page is shown below. The arrows have been added to show three prominent landmarks.
We download the image seen above (using a menu on the web page that is not visible in the screenshot) as palo_alto.jpg and we also download the associated world file using the name palo_alto.jpgw.
Note the features shown by arrows in the illustration. We would like to download an adjacent image showing the region just below the region seen above. To do so, we pan the image downward by clicking on the "South" arrow at the bottom edge of the image on the Terraserver page.
Terraserver will scroll down one image at the current zoom level and show the display seen above. The same three landmarks are marked again with arrows. Note that when Terraserver scrolls to a new adjacent image it leaves some overlap with the previous image. We download this second image as palo_alto2.jpg and the associated world file as palo_alto2.jpgw.
If we take a look at our downloaded files with Windows Explorer we see that we have downloaded four files: two .jpg image files and two .jpgw world files. World files are simple text files that may be opened with Notepad.
Step 2: Import and georegister the images
Unfortunately, world files are obsolete in GIS usage because they do not encode the projection used. They simply provide certain local projection values and leave it up to the user to manually specify the actual projection in use.
If we opened up one of the world files in Notepad we would see the following for the palo_alto.jpgw:
8.000000
0.000000
0.000000
-8.000000
574400.000000
4147200.000000
....and this for palo_alto2.jpgw:
8.000000
0.000000
0.000000
-8.000000
574400.000000
4145600.000000
… That's all! No mention of any projection used, just the local projection parameters for whatever mystery projection was used. To find the projection used we will have to drill into the documentation posted in obscure places on the TerraServer site and then use the information we find to manually specify the projection information for each image within Manifold.
This is a phenomenally inconvenient way of doing things, which is why the GIS industry evolved the GeoTIFF standard that is now used by over 160 different companies instead of world files. However, Microsoft's TerraServer still uses world files. One suspects they do so because using GeoTIFF would force them to provide the images in TIF format while using world files allows them to serve images in JPG format and thus simplify their download process to simply saving whatever image has already been served to a web browser in JPG.
Be that as it may, if we want to use Terraserver images for free we must deal with world files. Lucky for us, Manifold is a powerful enough tool to cut through the hassles associated with world files. When Manifold imports an image it automatically looks in the folder from which the image was imported to see if there is a similarly-named file that uses one of the typical world file extensions. If such a world file exists, Manifold will extract the information it contains for use with local projection values. In this case, Manifold will find a .jpgw file for each .jpg file that contains the world file numbers. See the Importing Images topic for more information.
We import the palo_alto.jpg and palo_alto2.jpg image using File - Import - Image. If we open the Palo_alto image we see that Manifold has correctly imported the .jpg image we downloaded from the web site.
If we open the Palo_alto2 image we see that it too has been correctly imported.
In both cases, because the world files do not specify projections Manifold has imported both images using Orthographic projection. Because the use of Orthographic is a tip-off to Manifold that the image file format from which the images were imported does not provide full projection information, Manifold has raised an info bar on both images to warn us to verify the projection.
We can verify the projection and georegister the images at the same time, by clicking the info bar to launch the Edit - Assign Projection dialog for the image. We will have to manually specify the projection to be used since the world files do not do so.
The Assign Projection dialog is used to verify the projection if no changes need to be made or, if changes must be made, this dialog is used to specify the correct projection and parameters. We first click on the info bar for the Palo_alto image.
As seen above the dialog shows that the image has been imported using the default Orthographic projection used when no projection is cited by the format used. However, the Local offset and Local scale values have been correctly set to the values contained in the world file.
Check the Preserve local values box so that the local values imported from the world file will be retained when we change the projection to the correct projection.
We now must specify the correct projection. In this case, it is Universal Transverse Mercator - Zone 10 (N) and the datum to use is the North American 1983. See the notes at the end of this topic to learn how we know this magic information.
Hunt through the hierarchical list of projections in the dialog to find Universal Transverse Mercator - Zone 10 (N) and then choose this projection by clicking on it. Choosing a new projection would normally reset the values in the Local offset and Local scale boxes to the defaults for the new projection. Since we have checked the Preserve local values box the local values imported from the world file will not be changed.
After we press OK the image is georegistered. There will be no apparent change in the appearance of the images since georegistration in this case is simply a matter of telling Manifold how to interpret the coordinates it already has.
Repeat the above procedure to verify and georegister the Palo_alto2 image as well by using the Edit - Assign Projection dialog.
Step 3: Copy and paste to assemble a larger image
Now that the images are georegistered the rest is easy.
Open the Palo_alto image. This is the "upper" image. Also open the Palo_alto2 image. This is the "lower" image that overlaps the Palo_alto image along the upper edge.
With the focus on Palo_alto2, press CTRL-A or choose Edit - Select All to select all of the pixels. Press CTRL-C or Edit - Copy to copy all of the pixels.
Click on the Palo_alto image to move the focus there and press CTRL-V or choose Edit - Paste to paste the copied pixels into this image. Manifold will ask if we want the image enlarged since most of the pixels added by the pasted image fall outside the original Palo_alto image. Press Yes.
The result is that the Palo_alto image has been extended downward by the pasting of the pixels copied from the Palo_alto2 image. The pasted pixels remain selected after the Paste operation.
If we click Select None we can see that the two images have been assembled into a single, larger image.
Zooming into the image we see that the alignment is excellent. The green lines mark the boundary between the original image and the pasted pixels from the other image. When TerraServer images originating in the same USGS DOQ are pasted together the boundaries are usually invisible.
Notes
This procedure uses two images that are vertically adjacent. The same process works when more images than just two are assembled. Import the images, georegister each using the Edit - Assign Projection dialog and then copy and paste them together into a single large image. When working with many images or large images, use a machine with lots of RAM, fast processor and adequate free space on disk.
How on Earth did we know that the correct projection to use was UTM zone 10? This resulted from some detective work and from some pure guesswork based on previous experience with USGS data sets. The TerraServer website provides only the following remarks in a discussion of the USGS "DOQ" images from which the TerraServer images are derived.
"The standard DOQ from the U.S. Geological Survey is a black-and-white (gray-scale) or color-infrared image covering 3.75 minutes of latitude by 3.75 minutes of longitude. Thus, four such photos can be combined in a mosaic to cover the area represented by a standard USGS 7.5-minute, 1:24,000-scale topographic map. Using mosaics is easier because images overlap. The DOQ's are referenced to the North American Datum of 1983 and use the Universal Transverse Mercator projection."
The datum in use is obviously the North American 1983 datum, most likely that for CONUS (meaning the Continental US, since that's where Palo Alto is located), but the projection is not as explicit as we would like. The Universal Transverse Mercator projection is not one projection but rather a family of over one hundred projections. Saying an image is in UTM without saying which specific UTM zone was employed is somewhat like saying that an address is located at a postal code without specifying which postal code. It's possible to do some detective work to figure out what postal code is involved for a given street and city but it would be nice to simply have the postal code specified.
It's a reasonable guess that if the DOQs from which the TerraServer images are generated are in some UTM projection than the TerraServer images are also in the same UTM projection. We can also infer that if the DOQs are arranged so that they tile standard USGS 7.5-minute, 1:24,000-scale topographic maps then the DOQs use the same UTM zone employed in USGS 7.5-minute topo maps. If we know the USGS topo map in which the TerraServer image is located, we know the UTM zone used for that image.
The Manifold CD contains a .map file called USGS 24K Index.map that provides an index to USGS 7.5-minute, 1:24,000-scale topographic maps covering the US. We can open this map and based on our knowledge of where Palo Alto is located (it is located conveniently between Mountain View and Menlo Park) we can see that Palo Alto, California, is contained within the Palo Alto USGS topo map. If we didn't know where our images were located we would have to do some additional detective work, perhaps consulting an online atlas or other resources to figure out in which USGS topo map they reside. Using Manifold, we might add a layer of place names to the USGS 24K index map. For the images used in this example the task is easy but in cases where TerraServer images are in rural areas near the edges of topo maps we may have to work harder to figure out in which topo map our images are located.
Once we know the topo map in which our images are located we know the UTM zone they are in. Somewhere on Internet there is probably a list that specifies what UTM zone is used for each USGS topo map. However, there is a faster way to find out what UTM zone our images are in than to hunt all over Internet for some magic list of USGS topo maps and the UTM zone used for each.
We can simply go to the USGS web site and download a 1:24K scale Palo Alto map in SDTS format (USGS calls it an "SDTS DLG" map). We import it into Manifold, open it and look in the Edit - Assign Projection dialog to see what UTM zone is used. Because SDTS format correctly saves full projection information the UTM zone used will be correctly stated. We will see from the dialog that it is UTM zone 10.
It's usually a good idea to download the USGS 1:24K SDTS DLG drawings for the region covered by the images being pasted together. The transportation / roads drawings from the SDTS map can be shown together with the images in a map to verify that georegistration has been correctly accomplished, as seen above.
One other use of copy and paste is to eliminate the small USGS logo that is embedded in each image downloaded from Terraserver. By pasting an adjacent image to the South the region of overlap will cover up the USGS logo with non-logo pixels. This effect can be seen in the example above where the USGS logo in the lower right of the Palo_alto image was covered up by overlapping pixels pasted from the Palo_alto2 image. The image can then be cropped back to the original size and the USGS logo will be gone. The images provided by USGS are all in the public domain and do not need to be branded with the USGS logo if the user does not so desire. The original USGS DOQ images do not include small logos.
See Also