A new method for image segmentation

摘要

A method is presented for finding a threshold surface whichinvolves the ideas used in other methods but attempts to overcome someof their disadvantages. The method uses the gradient map of the image topoint at well-defined portions of object boundaries in it. Both thelocation and gray levels at these boundary points make them good choicesfor local thresholds. These point values are then interpolated, yieldingthe threshold surface. A method for fitting a surface to this set ofpoints, which are scattered in a manner unknown in advance, becomesnecessary. Several possible approaches are discussed, and theimplementation of one of them is described in detail. Two versions ofthe C.K. Chow and T. Kaneko algorithm (1972) and the present algorithmare applied to a few images, and the latter is shown to give betterresults, matching human performance quite well