We can divide the dataset into subfiles, each containing only a single zone
and the immediate neighbors of readings in the zone. We further subdivide
these files into the connected components of the zone.
Each zone has an associated interval of values, given by
its critical value and its parents value.
For constructing the iso-surfaces at a fixed threshold
., we need only input data
from the zones whose intervals contain
.
In fact we can begin by reading
a single
data point from each component of a relevant zone. From each such point we can
quickly find a threshold crossing, searching within the zone.
We can then (using SpiderWeb) only input those points of the zone
that occur on hypercubes containing hits.
Obviously, whenever a hit occurs on a cube edge, all cubes sharing the edge also have hits.
For a single iso-surface, we need only input readings from within a single component of
a zone.
This is important as researchers have been concerned with preprocessing the data so that excessive I/O is not performed. One could further organize the readings in a zone using an oct-tree, an interval-tree, or a similarly appropriate data structure.