An optimal multiple threshold scheme for image segmentation

A fast search scheme is presented for finding single and multiple thresholds that maximize interclass variance between dark and bright regions. The original formulation of the problem by N. Otsu (see ibid., vol.9, no.1, p.62-6, 1971) had suggested an exhaustive search. The present method is orders of magnitude faster and uses a criterion function for its search scheme. The criterion function is derived by assuming a continuous probability function for the gray level histogram. Several examples are presented to illustrate the performance of the proposed scheme.