Optimizing raster storage: an examination of four alternatives.

The performance of many algorithms in spatial data processing depends on the way in which spatial entities are ordered. The paper defines a general class of problems in which the objective is to preserve as far as possible the spatial relationships present in two dimensions. Applied to a raster, the problem leads to the N or Morton ordering and a new Pi-order is proposed based on a familiar Peano curve. An algorithm is given for defining Pi-order. These are compared empirically and analytically to the conventional row ordering, and a simple variant, using a number of standard images and using a class of indices which includes the spatial autocorrelation measures. The empirical results support Pi-order but the analytic results are mixed.-Authors