Image Segmentation Using both Edge and Region Information
Takahiro Sugiyama, Keiichi Abe · 1994
This paper proposes a method for segmenting an image into only the regions segmentable reliably, using both edge and region information. It is important that lower level processes give useful information to upper level processes, for example, whether segmented regions are reliable or not. So we first modified the definition of region segmentation proposed by Horowitz and Pavlidis[41 and determined the condition of a certain region which can be reliably segmented. The condition with predicates utilizes both edge and region information. Our method segments only the areas satisfying the condition into the certain regions starting with the subspaces with size-scale s, leaving the rest of the image as uncertain regions.