Nouveaux filtres optimisés pour la détection de lignes de crêtes

This paper introduces two new linear filters to detect ridge lines. These filters are designed to respond to the requirements of a particular practical application to improve the working quality of a farm implement. The first filter is similar to Ziou’s filter and is implemented as a 2 order recursive filter. The second filter is designed following Canny’s methodology to optimally detect a “smoothed-exponential” edge whose profile is linked to experimental measurements relative to a practical application. The recursive implementation of this filter is of the 4 order. The edge detection results obtained with synthetic and real images demonstrate the superiority of the two filters presented in this paper with respect to Ziou’s filter for the targeted application. The second filter provides the best results in the case of noisy images.