Cluster counting for E799/E832 at Fermilab

A high-speed algorithm for counting the number of geographically contiguous above-threshold hits (clusters) in a CsI array is presented. The algorithm processes an N*N array in O(N) time, using O(N) parallel logic elements with only four inputs each. For the 40*40 (equivalent) array in the E799/E832 experiments at Fermilab, this cluster count information will be made available to the Level 2 trigger in less than 2 mu s. This approach is based on simple pattern recognition of local concavity and convexity in the signals from adjacent blocks in the CsI array. This approach is a two-dimensional analog to the Huffman-Clowes edge-labeling scheme developed for artificial intelligence and scene interpretation. The status of the implementation of this algorithm using Xilinx XC3042 field programmable gate arrays is presented.<>