Non-Parametric Adaptive Approach for the Detection of Dominant Points on Boundary Curves Based on Non-Symmetric Region of Support

In this paper, we propose an efficient and effective non-parametric method for detecting dominant points on a closed boundary curve. The proposed method determines a region of support which is not necessarily symmetric for a point on a curve without requiring any input parameter. The statistical and geometrical properties associated with the small eigenvalue of the covariance matrix of a sequence of connected points are explored for determining the region of support adaptively for a point. Once the region of support of a point is determined, the reciprocal of the angle made at that point due to its left and right arms is estimated as the curvature at that point and that estimated curvature signifies the point as a dominant point. The points that bear local maxima curvature are selected as dominant points. If two or more points are selected as local maximum points within small vicinity then the point which bears longer region of support is retained. The results of the experiments conducted reveal that the proposed method is highly consistent with human perception and outperforms other well known methods.