Corner detection by sliding rectangles along planar curves

A new corner detector for planar curves is presented in this paper. This algorithm finds corner by sliding set of three rectangles along the curve and counting number of contour points lying in each rectangle. This structure incorporates more than one view of given shape which is a key to find all corners successfully. Proposed technique was found very consistent with human vision system. This is an efficient method, as it does not involve calculation of cosine angle and curvature. Criteria for evaluation of corner detection algorithms are proposed in this paper. A comparative study of six corner detectors (including proposed algorithm) is also presented. This technique is very useful to detect corners from noisy shapes and natural object boundaries.