Hypercube algorithm for image component labeling

Labeling the connected regions of a digitized image is a fundamental computation in image analysis and computer vision. By assigning a unique label to each connected region, higher level image operations can identify, extract, and process different connected regions separately. Because of its primary importance, the problem has attracted research in developing parallel algorithms. Most of the research has been theoretical in nature, with notable exceptions. We present a new component labeling algorithm that is a parallelized hybrid of the sequential algorithms of R.M. Haralick and L.G. Shapiro (1979) and A. Rosenfeld, J. Pfaltz (1966). Experimental studies on the nCUBE/10 hypercube system at the University of South Carolina show that the algorithm has a relatively efficient balance of time complexity and storage utilization.<
>