Thresholding Using the Isodata Clustering Algorithm
MARYLAND UNIV COLLEGE PARK COMPUTER SCIENCE CENTER
Pagination or Media Count:
A recently proposed iterative thresholding scheme turns out to be essentially the well-known ISODATA clustering algorithm, applied to a one- dimensional feature space the sole feature of a pixel is its gray level. We prove that in one dimension, ISODATA always converges. We also apply it to requantize images into specified numbers of gray levels.