An efficient algorithm for smoothing, linearization and detection of structural feature points of binary image contours

In this paper, an efficient algorithm is developed for smoothing, linearization and detection of structural feature points of binary image contours. Using this algorithm, we can remove noisy pixels along a contour, decompose the contour into a set of straight lines, and detect structural feature points which correspond to convex and concave segments along the contour. Our algorithm only involves operations on contour chain codes and does not require float point calculations, thus it can be efficiently implemented. We have tested our method with handwritten character images. Copyright © 1996 Pattern Recognition Society. Published by Elsevier Science Ltd.