Paradigm for shape-based image analysis

Traditional image segmentation techniques typically divide an image into separate regions based on gray-scale characteristics. Most real-world image-segmentation problems, however, require some subsequent shape-based processing to yield acceptable results. Unfortunately, choosing an appropriate sequence of image-processing operators (a process) for this purpose can be a time-consuming, tedious procedure that requires considerable image-processing expertise. We describe a semiautomatic paradigm for selecting shape-based operations for an image-analysis process. Desired shape information for image regions is provided by the user in the form of easily specified cues. The cues are then automatically interpreted to select suitable image-processing operators and operator parameters; the operators can be morphological, topological, and image-manipulation functions. The paradigm, hence, enables easy prototyping of image-analysis processes for different problems. The user is not required to be an image-processing expert to apply this strategy - he or she need only be able to specify the desired shape properties of the regions in the image. We demonstrate our approach for both 2-D and 3-D image analysis problems.