AUTOMATIC EDGE DEFINITION BASED ON TOPOLOGICAL INFORMATION

摘要

This paper describes a new automatic edge definition algorithm. The process is divided into four stages: image transformation, edge pixel candidates definition, pixel classification, and edge linking. In the first stage the original image is transformed into general edges. Then a threshold is computed under a human visual perception point of view to define edge candidate pixels. The next stage is to classify the pixel candidates as edges / no edges using topological information. The last step is to link the edges. Besides this is an automatic edge definition algorithm, and generates an image with one pixel wide, results shown that the proposed algorithm defines better edges than other techniques as shown in this paper.