CLUSTERING EDGE VALUES FOR THRESHOLD SELECTION

Thresholds may be chosen for images containing several object classes by clustering thinned edge points in a 2-D histogram, whose axes represent gray level value and edge value. Each such edge cluster suggests its average gray level as a threshold. Interior clusters may also be defined as representatives of object class interiors. The relation of edge clusters to interior clusters gives rise to a classification strategy based on partitioning the 2-D histogram into disjoint regions labeled as to object class. Each partition is a classification domain for points of the original gray level image. © 1979 Academic Press, Inc.