Understanding Raster Data

 

Raster datasets are data arranged as pixels, in rows and columns. Each pixel has a value, and this value can represent more or less anything, from elevation (in a DEM) to a grayscale value (in an image) or anything else in between.  Here we will just concern ourselves with grayscale and color images, such as aerial photography, satellite imagery, hillshade maps, etc., that might be used as background image information for mapping using ArcPad and the iPAQs.

 

In general, there are two types of imagery that we will need to understand, grayscale images and color images. Of these two, grayscale images are by far the simplest.  Grayscale raster datasets consist of agrid of cells that contain values between 0 and 255, where 0 is black, 255 is white, and each number in between is an increasingly lighter shade of gray. Grayscale images are known as single-band images, since all of the information necessary to render an appropriate grayscale image is contained in a single grid of pixels.

 

Color raster datasets, on the other hand, can be represented in one of two ways, and understanding how to differentiate between these two types, and how to work with each, is important for manipulating and cropping images for use with ArcPad.

 

Multi-band Images

 

Multi-band images, as the name implies, are composed of multiple layers of pixels, typically three. One band each is used to describe the color intensity of red, green, and blue at a specific pixel location in a color image.  The pixel values are often written as(R, G, B), where R is the red value, G is the green value, and B is the blue value. For example, a black pixel in an image would have a value in the red layer of 0, in the green layer of 0,and in the blue layer of 0, or (0, 0, 0). A white pixel would be (255, 255, 255), while a red pixel would be (255,0, 0) and so on. This type of image is typically used for imagery with many colors, such as aerial photography or satellite imagery. If one were to import a multi-band image entitled color image to an Arc grid called colimg, three grids would actually be created, colimgc1 for the red band, colimgc2 for the green band, and colimgc3 for the blue band. Collectively, these grids will be linked as a stack with the name colimg, but no grid operations can actually be performed on the stack. Therefore, if you wished to crop the image color_image, you would have to import the color_image to the grid stack colimg and clip each of the three bands. Instructions for this can be found here.

 

Indexed Images

 

 Indexed images provide a means to store large images with a limited color palette in a relatively small amount of space.  Instead of three full bands, each containing a value between 0 and 255, an indexed image contains a single band containing a small number of values, usually ranging from 0 up to some limited number. In addition to this band, the image also contains a look-up table that maps each value to a specific RGB color (for example, the cell value Ô0Õ may represent the RGB color (0, 0, 0), while the cell value 5 represents the RGB color (123, 55, 42).  This type of raster image is highly efficient for images with a limited number of colors, such as USGS digital raster graphics (13 colors).

 

Indexed images are wonderful for display in ArcMap and ArcView, as they are natively supported by these programs, which are capable of interpreting the embedded color index.  Unfortunately, indexed images are not natively supported in ArcInfo, as it does not interpret the embedded color information.  This problem is a double-edged sword. Since indexed images are single band, they are much easier to clip in ArcInfo than multi-band images.  To view the images for actual clipping however, the color map of the image must be available as a separate file, with the extension .lut (an ArcInfo lookup-table; such tables for USGS DRGs are available. If such a lookup-table is not available, then Photoshop can be used to convert the indexed image to a multi-band image (RGB image in Photoshop terminology). This process will not affect the georeferencing of the new image, as long as the old world file (e.g. .tfw file) is renamed to match the new image. At this point, the new image can be treated as a multi-band image.