Computer vision using quadtree refinement (artificial intelligence, segmentation, knowledge)
Lewis W. Tucker · 1984
A computer vision system is described which utilizes a hierarchical search strategy to perform image segmentation by successive refinement of a quadtree. The quadtree, (a tree data structure with either 0 or 4 descendants from each node), is shown to provide a convenient mechanism for storing, in a topographical manner, semantic information as well as image data. Procedural rules associated with model elements drive the system in a cycle of hypothesis generation, scene evaluation, and quadtree refinement. Model-derived knowledge is incorporated into object experts which exercise control over the refinement process, concurrently performing segmentation and scene analysis.