We have developed a semi-automatic image segmentation tool which
combines conventional manual segmentation utilities with a novel
automatic image segmentation algorithm.  Manual segmentation is
achieved by dropping control points and fitting cubic splines to these
points.  Automatic segmentation is achieved by bimodally thresholding
local windows of the target image and countour following.  By
combining these two segmentation methods, a user can obtain acurate
boundry descriptions with much less effort.