A new heuristic edge extraction technique
J. Shu · 2005
The problem of detecting intensity changes within images is canonical in computer vision. This research paper presents a new heuristic edge extraction (HEE) technique for solving the problem. The HEE consists of the following processes: (1) preprocessing, (2) Sobel edge detection, (3) thinning edge operation, (4) AI A* algorithm (minimum-cost-searching scheme), and (5) chain coding. The preprocessing improves image quality to facilitate edge extraction. Sobel edge detection is used for obtaining the spatially accurate multipixel-wide edges within input images. The thining edge operation yields one-pixel-wide edges by processing the Sobel edges. AI A* algorithm extracts optimal curve edge segments based on the use of both thinned edge intensity and Sobel edge direction (i.e. gradient direction). Finally, the chain coding technique is applied to encode extracted optimal curve edge segments. The HEE provides, therefore, spatially accurate, high-quality, and exactly one-pixel-wide edges as system output.