Pairwise matching of 3D fragments

We propose a fractured surfaces matching algorithm for reassembling broken solids. The algorithm first use a multi-scale curvedness-based method to select feature points and remove less reliable feature points according to confidence value. Subsequently we obtain matching point pairs through measuring the similarity of local surface patches which are described by 3D histogram. After that we employ a fast and efficient hierarchical greedy algorithm to align the two fractured surfaces and judge if they are matched according to surface distance. The algorithm is also suited for arbitrary surface matching which contain rich feature points.