Thresholding Using the Isodata Clustering Algorithm
F. R. D. Velasco · 1979
An investigation is made of the use of the ISODATA clustering algorithm as applied to a one-dimensional feature space. For two classes, the ISODATA turns out to be an iterative thresholding scheme, which is very convenient for its simplicity. In this case (two classes)it is proved that the ISODATA algorithm always terminates. For a number of classes larger than two, ISODATA can be used to requantize images into a specified number of gray levels.