Deformable model-guided region split and merge of image regions

摘要

The incorporation of model-based splitting operations produces an improved algorithm for image segmentation with deformable templates. Image regions are merged together and/or split apart, based on agreement with a statistical shape model, as well as local shape properties of the region bounding contours. Experiments indicate that the new split/merge strategy yields a significant improvement over the previous template-based segmentation method that used merging alone.