These two papers propose clustering as a deterministic alternative to RANSAC. The corresponding pairs of points can be rewritten as points in 2D space. The inliers will be found close to each other while the outliers will be found further away from the cluster, depending on how well the points correspond. The first paper describes the idea and shows it can be used for outlier removal. It can handle both scale and rotations, but seem to be more robust when that is not the case.
- Clustering in 2D as a Fast Deterministic Alternative to RANSAC. A. Hast, G. Kylberg. The second workshop on Features and Structures (FEAST), colocated with ICML, Poster. pp. 1. 2015.
@inproceedings{Has15a, author = {A. Hast and G. Kylberg}, title = {Clustering in 2D as a Fast Deterministic Alternative to RANSAC}, booktitle = {The second workshop on Features and Structures (FEAST), colocated with ICML}, pages = {1}, note = {Poster with Paper}, year = {2015}}
By setting the threshold loosely it can be used as a preconditioner to RANSAC, removing the majority of the outliers. This can be especially useful when they are many as RANSAC performs poorly when they are more than 50% of the points.
- An Efficient Preconditioner and a Modified RANSAC for Fast and Robust Feature Matching. A. Hast, A. Marchetti. International Conferences in Central Europe on Computer Graphics, Visualization and Computer Vision (WSCG’12), Communications Paper. pp. 11-18. 2012.
@inproceedings{Has12a, author = {A. Hast and A. Marchetti}, title = {An Efficient Preconditioner and a Modified RANSAC for Fast and Robust Feature Matching}, booktitle = {International Conferences in Central Europe on Computer Graphics, Visualization and Computer Vision (WSCG 2012)}, pages = {11--18}, note = {Short Paper}, year = {2012}}
Anders Hast 