The "R" Word
What do most of the leading desktop mapping systems on the market today have in common? They don't provide a convenient and correct method for building, and projecting raster image maps. Unfortunately this has only added to the tendency of the average desktop mapping user to generally avoid raster data as something that seems too difficult, and/or foreign for them to work with. With the tremendous advances in scanning, digital orthophotography and satellite imaging technology, most desktop mapping users are missing out on a relatively inexpensive source of rich, visual data - raster data - that could significantly improve their spatial decision making.
      Why has raster become the "R" word to the vast majority of desktop mapping users, and in fact most GIS professionals? Like so many issues, in general, it's a lack of training and/or education, and from a practical point of view, a lack of available, easy to use software tools. There is no question that working with raster data can be intimidating, if one approaches it from a theoretical point of view, but how many people really understand topologic data structures, polygon processing and other vector data handling concepts? In most cases, the majority of desktop mapping users simply rely on their software to manage these technical issues for them. They could do the same with raster data.

A Component Approach
To date however, desktop mapping users have not had access to raster software tools of the same quantity and quality as their vector counterparts, but things are beginning to change. With the increasing performance of PCs, the leading raster data handling packages are all migrating their systems from their Unix workstations to the PC platform.
      In addition, new products are entering the market, designed from scratch for PCs, and therefore tending to be easier to use, and more focused. This new class of geographic software is designed to work seamlessly with existing desktop mapping systems, complimenting them by adding missing functionality.
      These new tools usually do not try to be all encompassing, preferring to take what might be called a component software approach in which a product is designed to perform a specific, limited task, as part of an overall "plug and play" strategy that provides customers with the flexibility to select only those tools which provide the necessary utility.
      As an analogy, the reader might compare this kind of scenario to the stereo music industry, in which customers are able to select the best component, in their price range, to provide a particular function, from companies that specialize in a particular technology. The key to the success of this strategy is standards, which thanks to Mr. Gates are now available, albeit under the control of a commercial venture - Microsoft.

Raster Image Map Basics
As noted above, most of the leading desktop mapping systems have very limited raster data processing capabilities. For example, in order to import a raster image into ESRI's ArcView or MapInfo the actual raster image scan file must be accompanied by a second file which contains the information needed to properly display the raster image in one of these desktop mapping system environments. In the case of ArcView this file is called an "ESRI-compatible world file," and in MapInfo the file is called a raster image table file. These "support" files provide the information needed to register and display the raster image map in each of these systems.
      In order to create these support files, a "raw" scanned image file, such as a TIFF or Bitmap file, must undergo a two-step transformation process. The first step in this process is to establish a real world coordinate reference for the image pixels. The second step is to remove the skew and distortion inherent in any raw scanned image by resampling each pixel and adjusting it such that it projects correctly in the coordinate system of choice. This skew and distortion can either be present in the original source image itself, or it can be the result of the scanning process, as it is generally not practical to exactly align the scanning grid with the image grid.
      In a raw scanned image each pixel is assigned a Cartesian coordinate pair based on an arbitrary row and column numbering scheme derived from the scanning grid and accompanying scanning software. Usually the pixel in the lower left hand corner of the image is designated (0,0) giving all other pixels in the image a positive coordinate greater than zero, using a standard X/Y coordinate system numbering scheme.

Georeferencing
In order to convert from "pixel coordinates" to "real world coordinates" the image file must be geographically referenced, or "georeferenced" to a minimum number of ground control points. The minimum number required is based on the mathematical model which the user selects. The more sophisticated the mathematical fitting scheme, the more control points that are required. A least squares method of solution is generally used to perform the transformation.
      Theoretically, the image could be georeferenced with just two points, but in most cases the minimal approach is to use the "affine" method, which requires at least three. In general, the more control points the better the solution. A first order polynomial fit requires a minimum of five control points, and a second order polynomial solution requires a minimum of seven points.
      The control points themselves are exact locations which must be able to be identified in the image, and on the ground. By establishing both the pixel and real world coordinates for a series of points, a mathematical model can then be derived, which in turn can be used to generate real world coordinates for all of the pixels in the entire image. Real world coordinates can be determined from on the ground surveys, perhaps utilizing GPS, or from information supplied with the original source document, such as grid or tick marks on a USGS quadrangle sheet.
      In order to check the relative accuracy of the solution the least square residual errors can be observed. Residual errors indicate the difference between computed and observed values. Adjustments in the georeferencing procedures can be made if it is determined that the residuals are not acceptable. Additional control points can be added, or those selected can be adjusted in order to arrive at an optimal solution.

Coordinate Systems and Map Projections
Once each pixel in the image has been georeferenced, the second step in the transformation process can then be undertaken. In this procedure the skew and distortion, inherent in any scanned image, are removed by resampling the image pixels and "projecting" them into a coordinate system in which the coordinate axis are orthogonal, or at right angles. This is required in order to provide the desktop mapping system with an image base map which has the same orientation and internal relationships as the vector data which it may be eventually compared with.
      As the pixels are being adjusted the user has the option to select which map projection system they want the final image to be displayed in. A unique relationship will always exist between the source image coordinate system and the destination coordinate system. For instance, the real world coordinates used during the georeferencing process may have been based on WGS 84, and the user may want to display the raster image base map in a 1927 state plane coordinate zone. The coordinate transformation from WGS 84 to the 1927 state plane zone is a standard conversion which can be applied during the processing.

Raster in the Background
Perhaps the most powerful use of raster image maps is found in their use as a background, or base map upon which vector data is overlayed. This combination offers desktop mapping users potentially the best of both the raster and vector worlds. With raster data in the background, users are able to "see" so much more than they could ever afford to have captured in vector form, yet they can still have the speed and power of vector data handling and processing for those details of interest.
      In general, raster image maps look more familiar to non-mapping professionals, and hence they provide these users with the ability to relate much more quickly to a desktop mapping environment. In order for the use of GIS technology to become as widespread as many people have been predicting, the cost of acquiring data, and the ability for people to relate to what they are seeing is going to have to improve significantly. The use of raster data offers significant advantages over vector in both of these areas. In the past, the PC platform has been challenged by the size of most raster images, but we all know what direction that concern is moving in.

Exciting New Raster Data Sources
With the recent decision to open up the use of high resolution satellite imagery the availability of raster data is going to take a giant step forward. Over the next few years this kind of data will become a commodity, with its use certain to radically change the economics of building and maintaining a land base. It may not be possible to predict exactly how all of this is going to shake out, with new consortiums being announced on a regular basis, but regardless of who the players are, it seems fair to conclude that the availability of high quality raster data can be taken as a given. The only remaining issue is whether the average desktop mapping user can overcome their fear of the "R" word, and position themselves to take advantage of this rich source of spatial information.

About the Author:
Gene V. Roe, Ph.D. is the director of business development for Blue Marble Geographics in Gardiner, Maine. He may be reached at 207-582-6747.

Back