next up previous
Next: Pre-Processing the ERDAS/Lan Up: Parallel Hoshen-Kopelman Algorithm Title Page Previous: Input File Format:

1. Hoshen-Kopelman Algorithm

A thorough discussion of an optimal implementation of the Hoshen-Kopelman (Hoshen and Kopelman, 1976) cluster identification algorithm is presented below. Particular emphasis is given to the input datasets, programming data structures, the cluster neighborhood rule, and the use of a finite-state-machine (FSM). The FSM discussion focuses on the three major implementation components: the temporary label assignment, the search path compression, and the formal finite state machine.

Michael W. Berry (
Sat Mar 30 23:40:13 EST 1996