Peter Meer (Electrical and Computer Engineering Department
A new robust estimation algorithm, the generalized projection based
M-estimator (gpbM) is proposed. The algorithm is general and can handle
heteroscedastic data where every point in the estimation have a different
covariance. Does not require the user to specify any (scale) parameters,
and can be applied to multiple linear constraints for single and
multi-carrier problems. The gpbM has three distinct stages: scale estimation,
robust model estimation and inlier/outlier dichotomy. The model estimation
stage can be further optimized by using Grassmann manifold theory. For
data containing multiple inlier structures the estimator iteratively determines
one structure at a time. We present four homoscedastic and five
heteroscedastic computer vision problems with single or multiple carriers.
Work done together with Sushil Mittal and Saket Anand.