Robust comparison of binary images

The Hausdorff distance between planar sets of points is known to be a good method to compare binary images. We present here a modification of this distance, called the censored Hausdorff distance. It allows robust comparisons of noisy binary images. The definition of the Voronoi surfaces is extended to this new distance, as well as the definition of the inclusion field. We then show that it is possible to localize a small binary pattern in an image.