4 - Spécification d'une ondelette pour l'analyse multirésolution d'un contour discret

A complex contour can be approximated at different scales of resolution to obtain a simple description containing less information but retaining enough information to permit recognition . We propose a multiresolution approach using wavelets to approximate a discrete contour represented by Freeman code. After specifying a modified wavelet for the problem, we present an algorithm consisting of a quantification phase and synthesis phase . Results of implementation of our algorithm are presented. At each resolution level, we compute the entropy of the discrete contour and the quadratic mean square error of the reconstructed contour is evaluated using the signal to noise ratio.