Kil-Moo Lee(1), Peter Meer(2) and Rae-Hong Park(1)
(1)Department of Electronic Engineering, Sogang University
C.P.O. Box 1142, Seoul 100-611, Korea
(2)Department of Electrical and Computer Engineering
Rutgers University, Piscataway, NJ 08855, USA
Robust high breakdown point estimators are now routinely employed
in range image segmentation algorithms.
We propose a novel technique using the adaptive
least k-th order square (ALKS) estimator which
minimizes the k-th order statistics of
the squared of residuals. The optimal value of k is determined
from the data, and the procedure does not require that the
pixels of the homogeneous surface patch are in
absolute majority in the window of analysis.
The ALKS shows a better tolerance to structured outliers
than other recently proposed similar techniques: MINPRAN and RESC.
The performance of the new, fully autonomous, range image
segmentation algorithm compares favorably to other methods.
This is an extended version of the correspondence published in,
IEEE Trans. Pattern Anal. Machine Intell.
vol. 20, 200-205, 1998.
Return to Research: Robust analysis of visual data
