A fast binary-image comparison method with local-dissimilarity quantification

Image similarity measure is widely used in image processing. For binary images that are not composed of a single shape, a local comparison is interesting but the features are usually poor (color) or difficult to extract (texture, forms). We present a new binary image comparison method that uses a windowed Hausdorff distance in a pixel-adaptive way. It enables to quantify the local dissimilarities and to give their spatial distribution which greatly improves the dissimilarity information. Combined with a support vector machine classifier, this method is successfully tested on a medieval-impression database