Morphological algorithms for centroid normalization in relational matching

The effect of iterative morphological filters on the determination of centroids is examined. On sample binary images that are subjected to high-frequency variations on their boundaries, increasing iterative sequences of morphological openings and closings are observed to normalize the centroid position. When grosser (or more structured) distortions are introduced to the boundaries, the notion of a normalized centroid is less meaningful. In these cases, it is interesting to consider measurements on the charging centroid trajectory over the iterative filters as an alternative structural signature. In addition, morphological opening and closing distributions can serve as tools in structural matching