Mapping Image Data

Set Up 1: Getting the Tutorial Data

1. In the Windows Explorer, navigate to the network drive  K: (aka \\Software\Winsoft), open the folder  Maps, and then open the folder  Introduction to GIS.
2. Drag the folder  mappingimages and its contents to either:
1. your network drive  U:, e.g. into the folder  My Documents; or
2. the local hard drive  C:, e.g. onto your Desktop.

Set Up 2: Initializing ArcMap and Adding Data

1. Start up the  ArcMap software (see Constructing and Sharing Maps for details).
2. Click on the button  Add Data.
3. In the dialog Add Data, navigate into the folder  mappingimages; if necessary, make a new connection to it (see Constructing and Sharing Maps for details).
4. In the folder  mappingimages, click on the file  amherst_2004.sid and then ctrl-click on the files  q117894.tifand amherst_boundary.lyrto add them to the selection. The first is an orthophoto of Amherst, and the second is a topographic quadrangle, a scanned map prepared by the United States Geological Survey.
5. Click on the button Add.

   

Procedure 1: Determining the Characteristics of a Raster Image

1. In  ArcMap, in the Table of Contents, double-click on the raster of interest, e.g.  amherst_2004.sid or  q117894.tif.
2. In the dialog Layer Properties, click on the tab Source.
3. Read the table Property | Value:
1. In the section Raster Information, you should note the following:
• The number of Columns and Rows in the raster (in this example, they are equal so it's square)
• The Cellsize (X, Y) (pixel size) in the units of the coordinate system (in this example, it's again square).
• The file Format;
• The Number of Bands per pixel;
• The Pixel Type and Pixel Depth;
• If a Colormap is used;
• If a NoData value is assigned;
• If Compression is used.
2. Scrolling down to the section Spatial Reference, you should note what that is, and also its Linear Unit (if it has one).
3. Scrolling up to the section Extent, note the distance the raster covers in each direction

Procedure 2: Georeferencing a Scanned Map

For this procedure you must already have a scanned map available, e.g. the 1833 map of Amherst shown above. You must also have some reference layers for comparison, such as boundary files, orthophotos, or GPS points.

1. Begin by adding the reference layer(s) and scanned map to ArcMap:
1. Add one or more reference layers for comparison, e.g.  amherst_boundary.lyrand  amherst_2004.sid (see Constructing and Sharing Maps for details).
2. If you know or can guess the projection of the scanned map, change the spatial reference of the map to match (see Mapping Geographic Coordinate Data for details). Otherwise, if you don't want to match the reference layer(s), a good option is Mercator, since it is shape-preserving and also orients north upward, a common characteristic of paper maps.
3. Add the scanned map, e.g.  amherst1833.sid.
4. In the dialog ArcMap, you will be advised that One or more layers is missing spatial reference information…; click on the button OK.
5. Because the scanned map has no spatial reference information, it will be positioned at the origin of coordinates, typically far from the reference layer(s).
• Optional Step: In the toolbar Tools, click on the button  Full Extent. Viewing the full extent of the data will likely produce two widely separated specks, one the correctly positioned reference layer(s) and the other the unplaced scanned map. Can you tell which is which?
6. To view the scanned map, right-click on its name in the Table of Contents and then click on the menu item  Zoom To Layer.
7. Examine the added map and get a good idea of its extent and any marked boundaries.
8. Return to the original location by right-clicking on a reference layer's name in the Table of Contents and then clicking on the menu item  Zoom To Layer.
9. Zoom in or out from the reference layer so that its recognizable features roughly match those of the scanned map.
2. Now initiate the georeferencing process:
1. If the Georeferencing Toolbar is not already visible, click on the menu View, then point at the menu item Toolbars, then click on the menu item Georeferencing. After the toolbar appears, you can dock it out of the way, by clicking-and-dragging it anywhere around the window frames.
2. In the toolbar Georeferencing, click on the menu Layer:, then click on the menu item for the scanned map (if isn’t already selected — ArcGIS will list all image layers without a spatial reference, and more than likely this will be the only one).
3. Click on the menu Georeferencing, and then click on the menu item Fit to Display. The result will look something like the image at the right.
4. This is a good time to save your map; in the toolbar Standard, click on the button  Save.
3. You must now add a control point that links the same recognizable location on the two layers, by first clicking on it on the scanned map, and second clicking on it on the reference layer.
1. Locations on the scanned map are recognizable in a number of ways:
• Point features are typically labeled;
• Linear features such as streets, railroads, rivers, canals, and political boundaries are usually labeled and have intersections or sharp corners;
• Survey markers will often have explicit coordinates printed next to them;
• A graticule will have intersections of meridians and parallels and explicit coordinates at the map edges.

In the last two cases it's usually easiest to guess a coordinate location on the reference map and then correct it later, as described below. Warning: to use such coordinates you must be working in the spatial reference of the scanned map!

2. When you have identified a location on both maps, in the toolbar Tools, click on the button  Zoom In, and then click and drag across both layers to draw a rectangle containing this location on both maps.
3. If you can't clearly distinguish this location on the scanned map, drag another small rectangle around it to zoom in further.
4. In the toolbar Georeferencing, click on the button  Add Control Points.
5. In the scanned map, click on this recognizable location.
6. If you’ve made a mistake, you can hit the key Escape to stop the link, and then continue with Step (i).
7. If you zoomed in a second time in Step (b), then in the toolbar Tools click on the button  Go Back To Previous Extent.
8. If you can't clearly distinguish the recognizable location on the reference layer:
1. In the toolbar Tools, click on the button  Zoom In, and drag another small rectangle around it to zoom in further.
2. In the toolbar Georeferencing, click on the button  Add Control Points. Notice that it still remembers that you have already initiated a control point by clicking on the scanned map.
9. In the reference layer, click on the recognizable location.
   
   The scanned map will now shift its position to bring the two points into alignment.
10. In the toolbar Tools, click on the button  Go Back To Previous Extent to return to the overview.
4. Repeat Step 3 with a second recognizable location; this will uniformly scale and rotate the map to align both the first and second points.
5. Repeat Step 3 a third time using a point that's widely separated from the line connecting the first two points. This will nonuniformly scale the map and rotate it to align all three points. This is called a first-order polynomial (affine) transformation.
6. After a fourth application of Step 3, most likely the two points linking the ends of the control point will no longer be perfectly aligned, having some residual distance represented by a blue line, as seen to the left. This is because there are no additional free parameters in this transformation, and a best fit must be calculated.
   
7. For most applications you will want to repeat Step 3 several more times, using points around the edge and then throughout the middle of the area of interest.
8. A full description of the control points you have set up is provided in the Link Table.
1. In the toolbar Georeferencing, click on the button  View Link Table.
   
   The dialog Link Table should now appear, listing each control point link and their starting (Source) and ending (Map) locations.
2. If you click on any control point link in the table, it will also be highlighted in yellow on the map.
3. The link table provides information about the residual distance between between the two ends of a control point link, and the Total RMS Error, an average of the residuals, which describes how far out of alignment the entire transformation is. We would like it to be as small as possible. Comparing individual residual distances to the total RMS error can indicate which control points are unusually separated. This might be due to:
• poor surveying;
• rerouting of roads, railroads, or canals, and the meandering of rivers;
• deliberate abstractions, e.g. the separation of features to make them more distinguishable;
• bad GPS readings;
• accidental clicks.

These points can be removed from consideration by clicking them in the table and pressing the key Delete.

4. The X and Y values in the Link Table are editable; this is most useful if the control points are survey markers or graticule intersections whose values are printed on the map, and can be typed into the fields XMap and YMap.
5. Warning: ArcMap does not store information about the link table, so to be able to return to where you left off after quitting or to restore from a crash, you should periodically save your table by clicking on the button Save… . This will let you create a text file storing your control points that can be reloaded later by clicking on the button Load… .
6. Click on the button OK to dismiss the Link Table dialog.
9. Another way to improve the fit is to use nonlinear transformations. Their effect on the scanned map may not always be desirable (for example, you wouldn’t use them on a presumably accurate map that merely needs to be positioned). There are several options available:
1. In the toolbar Georeferencing, click on the button  View Link Table.
2. In the dialog Link Table, click on the menu Transformation:, and then click on one of the following menu items:
• If you have at least six control points, the item 2nd Order Polynomial becomes available. With exactly six, the Total RMS Error will be zero.
• If you have at least ten control points, an additional option is 3rd Order Polynomial . With exactly ten, the Total RMS Error will be zero.
• Also available with at least ten control points is the option Spline. It provides an exact fit for all additional control points, but can be very slow due to the large number of calculations required.
• An option available at all levels is Adjust; it is very fast for even hundreds of control points, but produces discontinuities in the image at the points' exterior boundary.
3. Click on the button OK to dismiss this dialog.
10. Once you’re satisfied with the fit of the transformed map, you can save it as a new raster layer for later use. Be aware that this process can take a while for a large map.
1. It's a good idea to first save your control points as described in Step 8(c).
2. In the dialog Georeferencing, click on the menu item Rectify….
3. In the dialog Save as, in the text field Output Location:, click on the button  Browse and select the folder (not the file) where you want to save the new raster.
4. In the menu Format:, choose an output format; for scanned maps, JP2 or JPG is preferred, though PNG can also be good for relatively simple images. JPG is the most compatible with external applications and will typically produce the smallest files if you are willing to sacrifice image quality.
• If you choose JP2 or JPG, in the text field Compression Quality (1-100): type a value or leave the default (anything less than 100 will be lossy).
5. In the text field Name:, adjust the file name to be more descriptive, e.g. amherst1833rectified.jp2. Don't change the file extension here, use Step (d) instead. Warning: GRID-format names must have a base that's less than 13 characters long.
6. Click on the menu Resample Type:, and then click on one of the menu items Bilinear Interpolation or Cubic Convolution(better but slower). The option Nearest Neighbor is best only for categorical data.
7. The new raster’s cell size is initially based on that of the scanned map, and it's usually best to leave it at the default. You can, however, reduce the file size by increasing the cell size, by typing a new value in the text field Cell Size:.
8. The new raster will be a rectangle in the current coordinate system, and that means that areas outside of the transformed map will be set to NoData. By default this value will be 0 (black), but you can assign those pixels another value (e.g. 1 — white) by filling in the text field NoData as:.
9. Click on the button Save.
11. Now review the rectified image:
1. Add the rectified image to your map, e.g.  amherst1833rectified.jp2.
2. The NoData areas can be made transparent as follows:
1. Double-click on the name of the rectified image in the Table of Contents to bring up the dialog Layer Properties.
2. Click on the tab Symbology;
3. Click on the checkbox  Display Background Value: (R,G,B); leave the default color as No Color.
4. Click the button OK.
3. If you use a file format other than JPG or JP2 or PNG, ArcGIS automatically calculates the statistics of the colors in the image, and then uses them to provide what it thinks is a better color display. This is almost always incorrect for an actual image (as opposed to rasters describing quantities like elevation). To turn off the use of statistics for color display:
1. Double-click on the name of the rectified image in the Table of Contents to bring up the dialog Layer Properties.
2. Click on the tab Symbology;
3. In the area Stretch, in the menu Type:, select the menu item None.
4. Click the button OK.
4. in the Table of Contents, click off the checkbox next to the name of the rectified image, e.g.  amherst1833rectified.jp2. You can now see that the rectified image matches the scanned map, the reference layer, and the control points.