ABSTRACT
The previous chapter described some of the commonest data structures for the storage of raster data and emphasised that at least one of the reasons that they were developed was to produce smaller files for raster GIS layers. Smaller files mean that less disk space is needed to store them and that it will be quicker to transfer them between the disk and the memory of the computer. A small file size should also mean that operations on the layers, such as queries and overlays, should run more quickly, because there are less data to process. However, this will only be true if the query can use the raster layer in its compressed form, whether this be run length encoded or in a quadtree or in any other format. If this is not possible, then the layer would have to be converted back to its original array format (which would take time) and then the query run on this expanded form of the data. So, the data structures which are used for raster data not only have to produce savings in file size, but they must also be capable of being used for a range of GIS queries. To illustrate this, we will consider some simple raster GIS algorithms, and show how these can be implemented using the three raster data structures described in Chapter 7.
