• w = # inliers in data / # points in data
  • w is not well-known beforehand
  • n : # points needed for estimating a model are selected independently
  • wn: the probability that all n points are inliers
  • 1 - wn: the probability that at least one of the n points is an outlier
  • k : # iterations for RANSAC
  • p : desired probability that the RANSAC algorithm provides a useful result
  • 1 - p = (1 - wn)k

I have some questions


  • a data point which has been picked up once can be selected again in another iteration?
  • weighted data points V.S. performance