COMPACT METHODS FOR STORING RASTER DATA: -
When each cell has a unique value it takes a total of n rows X m columns X3 values ( X, Y coordinates and attribute value) to encode each overlay. If sets of cells within a polygon or a mapping unit all have the same value, however, it is possible to effect considerable savings in the date storage requirements for the raster data, providing of course that the date structures are properly designed. There are four main ways in which compact storage can be achieved.(i) Chain Code: - Chain code provide a very compact way of storing a region representation and they allow certain operation such as estimation of areas and perimeters or detection of sharp turns and concavities to be carried out easily. On the other hand overlay operations such as union and intersection are difficult to perform without returning to a full grid representation. Another disadvantage is the redundancy introduced because all boundaries between the regions will be stored twice.
(ii) Run-Length Codes- Run-length codes allow the point in each mapping unit to be stired per row in terms, from left to right, of a begin cell and an end cell. These are useful in reducing the volume of the data that need to be input to a simple raster database.
(iii) Block Codes- The data of run- length codes can be extended to two dimensions by using square block to tile the area to be mapped . The data structure consist of just three numbers, the origin (the centre or bottom left) and radius of each square. Both run-length and block codes are clearly most efficient for large simple shapes and so for small-complicated areas that are only a few times larger than the basic cell. For some operations data stored in block run-length codes must be converted to simple raster format.
(iv) Quadtree- The fourth method for more compact representation is based on successive division of the 2n X 2n array into quadrants are wholly content with the region. The lowest limit of division is the single pixel. The block structure can be described by a tree of degree 4, known as the quadtree.
If each cell represents a potentially different value, than the simple N X N structure is difficult to improve upon. Its limitations largely related to the volume of data and size of memory required. When� regions are presents, as is assumed to be considerably reduced by using length codes appear to be most efficient when the pixel size is large with respect to the areas of the region being displayed and stored, as resolutions improves and pixel numbers per region increase. However, block codes and quadtrees become increasingly attractive. The quadtree representations have the added advantage of variable resolution.
