Robust and Fast Edge Detection and Description in Range Images

A novel algorithm is presented for robust and fast edge detection and description in range images. In contrast to many other known methods, our algorithm provides a rich description of edge structures including detailed edge types and local surface characteristics. The proposed algorithm has been tested on a large number of real range images acquired by three scanners with quite different noise characteristics, and demonstrated good results. The computation time for one image lies between 1.5 and 10 seconds on a Sun Sparcstation 5, dependent on the image resolution. The highly parallel nature and the simple control structure of our algorithm makes a further significant speedup on a parallel architecture possible. no quantitative characterization of edge strengths [ l l ] , no straightforward geometric interpretation of the quantitative characterization of edge strengths [I, 51, no support of the classification in detailed edge