An adaptative method to smooth discrete curves proposed as a final step for edge detection

Discrete curves are usually smoothed either by approximation or by interpol ation. Our method leads to a solution, which can be gradually chosen from a spline interpolation to a rough approximation. A theoretical and an experi mental study both demonstrate the capability of one parameter to control th e behaviour of our smoothing algorithm. A deformation model is introduced t o optimise the choice of this parameter. Results are provided on sets of po ints corresponding to standard or noisy geometrical distributions. The algorithm is proposed as a final step in an edge detection process to g et a contour from a set of edge pixel candidates. (C) 2001 Elsevier Science B.V. All rights reserved.