Table 1.
The representative methods for point set registration.
| Research Study | Pairwise/Groupwise | Method | Rigid/Non-Rigid | Parametric/Non-Parametric Model | Characteristics |
|---|---|---|---|---|---|
| Besl and McKay [23] | Pairwise | Distance-based method | Rigid | Parametric Model | (1) Sensitive to the initialization (2) Trapping into local minima |
| Gold et al. [34] | Pairwise | Distance-based method | Rigid | Parametric Model | (1) Combining deterministic annealing and softassign optimization (2) Restricting to perform the rigid-body transformation |
| Chui et al. [35] | Pairwise | Distance-based method | Non-rigid | Parametric Model | Difficult to extend to perform higher dimension |
| Tsin et al. [36] | Pairwise | Distance-based method | Rigid and Non-rigid | Parametric Model | Maximizing the KC of point sets |
| Jian et al. [38] | Pairwise | Distance-based method | Rigid and Non-rigid | Parametric Model | Minimizing the Euclidean distance of two GMMs |
| Leordeanu et al. [46] | Pairwise | Distance-based method | Rigid and Non-rigid | Non-Parametric Model | Convexifying the QAP problem by spectral relaxation method |
| Cour et al. [47] | Pairwise | Distance-based method | Rigid and Non-rigid | Non-Parametric Model | Convexifying the QAP problem by semidefinite-programming relaxation |
| Almohamad et al. [50] | Pairwise | Distance-based method | Rigid and Non-rigid | Non-Parametric Model | Convexifying the QAP problem by doubly stochastic relaxation |
| Zhou et al. [22] | Pairwise | Distance-based method | Rigid and Non-rigid | Non-Parametric Model | Factorizing the large pairwise affinity matrix into some smaller matrices |
| Sandhu et al. [67] | Pairwise | Filter-based method | Rigid | Non-Parametric Model | Using a particle filter to register the point sets |
| Li et al. [16] | Pairwise | Filter-based method | Rigid | Non-Parametric Model | (1) Using a cubature Kalman filter to register the point sets (2) The correspondence should be computed in advance |
| Myronenko et al. [25] | Pairwise | Probability-based method | Rigid and Non-rigid | Parametric Model | (1) Using a GMM model to formulate the distribution of the point sets (2) Maximizing the likelihood of GMM |
| Ma et al. [76] | Pairwise | Probability-based method | Rigid and Non-rigid | Parametric Model | Developing a locally linear transforming for local structure constrict |
| Wang et al. [104] | Groupwise | Information theoretic measure | Rigid and Non-rigid | Parametric Model | Proposing a CDF-JS divergence as the cost function |
| Chen et al. [105] | Groupwise | Information theoretic measure | Rigid and Non-rigid | Parametric Model | Developing a CDF-HC divergence as the cost function |
| Giraldo et al. [106] | Groupwise | Information theoretic measure | Rigid and Non-rigid | Parametric Model | Using a Rényi’s second order entropy divergence as the cost function |
| Rasoulian et al. [108] | Groupwise | Probability-based method | Non-rigid | Parametric Model | Assumed that the multiple point sets are the noisy observations of mean point set |
| Evangelidis et al. [2,3] | Groupwise | Probability-based method | Rigid | Parametric Model | Assumed that the multiple point sets are transformed realizations of mean point set |