Image segmentation by syntactic method

摘要

A syntactic approach to image segmentation is proposed. Edge, line (curve), and corners are identified as image primitives. A set of ideal edges, lines and corners are used to infer a tree grammar. A weighted structure-preserved error correcting tree automaton algorithm is used to find the best match between ideal primitives and actual primitives in the image. Finally, an edge-linking procedure is used to link the edge-pixels.