Quite numerous computer vision applications require an assignment of the elements of a set of some extracted salient positions/descriptors to their corresponding counterpart in a model set, at some point of their proceeding. Examples are object recognition or image registration schemes. Because an exhaustive evaluation of all possible combinations of individual assignments is infeasible in terms of runtime, more sophisticated approaches are required, and some of them will be presented in this chapter. Possible ways of speeding up the search are applying heuristics, as done in the so-called search tree, or iterative approaches like the iterative closest point method. A typical challenge when trying to find the correct correspondences is the existence of a quite large number of outliers, e.g., positions being spoiled by gross errors. The straightforward approach of minimizing total deviations runs into difficulties in that cases, and consequently, methods being more robust to outliers are required. Examples of robust schemes are the random sample consensus (RANSAC) or methods transforming the problem into a graph representation, such as spectral graph matching or bipartite graph matching as done in the so-called Hungarian algorithm.
KeywordsDrilling Cross Correlation
- 3.Choi S, Kim T, Yu W (2009) Performance evaluation of RANSAC family. In: Proceedings of the British machine vision conference, London, pp 81.1–81.12Google Scholar
- 8.Harris C, Stephens M (1988) A combined corner and edge detector. In: Alvey vision conference, Manchester, pp 147–151Google Scholar
- 10.Jain V, Zhang H (2006) Robust 3D shape correspondence in the spectral domain. IEEE Int Conf Shape Model Appl 118–129Google Scholar
- 13.Leordeanu M, Herbert M (2005) A spectral technique for correspondence problems using pairwise constraints. In: Proceedings of the 10th IEEE international conference on computer vision, vol 2, Beijing, pp 1482–1489Google Scholar