Hausdorff distance map classification using SVM

We investigate a new pattern recognition technique, based on support vector machines (SVM). Our objective is to find in database of images constituted from wooden graven, the impressions which represent the same stamps thus illustrating the same scene. In this research, the statistical classification technique that includes Hausdorff distance, similarity measures and SVM has been developed for automatic distance maps classification. These distance maps are constructed at each scale by the computation of the Hausdorff distance between two binary images through a sliding-window. The efficiency of the proposed procedure is demonstrated in terms of classification rates, robustness and computing time at multi-scale resolution.