Fig. 1

High-level visualization of RASTER (best viewed in color), using a simplified example on a small 5 × 5 grid. Please also refer the section “RASTER” and Table 1. The precision of the input is reduced, with leads to an implied grid. This grid is shown to aid the reader but it is not explicitly constructed by the algorithm. The original input is shown in a, followed by projection to tiles in b where only significant tiles that contain at least \(\tau = 4\) values are retained. Tiles that contain less than \(\tau\) values are subsequently ignored as they are treated as noise. The result is a set of significant tiles. The parameter \(\mu\) specifies how many significant tiles a cluster has to contain as a minimum. In this case, given a minimum cluster size of \(\mu = 2\) and a maximum distance \(\delta = 1\), i.e. significant tiles need to be adjacent, two clusters emerge (cf. c), which corresponds to RASTER. Clusters as collections of points are shown in d, which corresponds to the variant RASTER′