Two graph searching techniques for boundary finding in white blood cell images

Two boundary finding algorithms based on a graph searching paradigm are applied to the problem of the accurate visual separation of touching white blood cells. Coded in FORTRANIV for the Nova 840 minicomputer, they produce very good results in modest running time. The algorithms are structured so as to cleanly isolate within a single mathematical function all problem-dependent information concerning the appearance of the boundary. For touching cells, this function takes into account the relatively high contrast in the vicinity of a point on the boundary, the difference in optical density between points in this neighborhood and an optimum derived from a histogram of the image, and the low degree of curvature the boundary is expected to exhibit. In principle, the information used may be quite diverse and adapted to the task at hand. As an example of their versatility, the methods are applied to the problem of finding the longitudinal axis of a chromosome using the geometry of the chromosome outline as a source of information.