Connected component labeling for binary images on a reconfigurable mesh architecture

摘要

We show how some existing component labeling algorithms for binary images could be speeded up by using the reconfigurable mesh architecture. Two algorithms are presented, the first one uses the ability of the reconfigurable mesh to perform certain operations in constant time, and the second one uses a bottom-up divide-and-conquer strategy. Both these algorithms have run times that are logarithmic functions of the image size.