Improvement of combinatorial Hough transform to compensate sampling errors of digital lines
Syuji Horiki, Takashi Watanabe, Akio Kimura, Norio Tayama · Systems and Computers in Japan · 1998
Conventional methods do not provide good polygonal approximation for contour images that are heavily disrupted by noise or other factors. Particularly, curved portions of a contour must be represented by a set of short line segments, but to extract such segments is a complicated problem under noisy environment. In this context, the authors proposed an improvement of the combinatorial Hough transform (CHT) to ensure polygonal approximation of discontinuous contours. The modified CHT employs adaptive voting to compensate sampling errors of digital lines. This ensures elimination of drawbacks that are inherent to conventional CHT or standard Hough transform (SHT). Evaluation experiments confirmed that the proposed method is efficient in polygonal approximation of complicated discontinuous contour images while being rather fast relative to CHT and SHT. © 1998 Scripta Technica. Syst Comp Jpn, 29(2): 57–67, 1998