Expected Performance of Robust Estimators near Discontinuities

Stewart, Charles V.

doi:10.1109/ICCV.1995.466829

Expected Performance of Robust Estimators near Discontinuities

Charles V. Stewart

ICCV 1995 pp. 969-974

doi:10.1109/ICCV.1995.466829 /iccv/1995/stewart1995iccv-expected/

Abstract

In extracting a polynomial surface patch near- an in-tensity or range discontinuity, a robust estimator must tolerate not only the truly random bad data (“random outliers”), but also the coherently structured points ((bseudo outliers”) that belong to a different surface. To characterize the performance of least median of squares, M-estimators, Hough transforms, RANSAC, and MINPRAN on data containing both random and pseudo outliers, we develop two analytical measures, ‘(pseudo outlier bias ” and ‘$pseudo outlier breakdown”. Using these measures, we find that each robust estima-tor has surprisingly poor performance, even under the best possible circumstances, implying that present es-timators should be used with care and new estimators should be developed. 1

ICCV Semantic Scholar

Cite

Text

Stewart. "Expected Performance of Robust Estimators near Discontinuities." IEEE/CVF International Conference on Computer Vision, 1995. doi:10.1109/ICCV.1995.466829

Markdown

[Stewart. "Expected Performance of Robust Estimators near Discontinuities." IEEE/CVF International Conference on Computer Vision, 1995.](https://mlanthology.org/iccv/1995/stewart1995iccv-expected/) doi:10.1109/ICCV.1995.466829

BibTeX

@inproceedings{stewart1995iccv-expected,
  title     = {{Expected Performance of Robust Estimators near Discontinuities}},
  author    = {Stewart, Charles V.},
  booktitle = {IEEE/CVF International Conference on Computer Vision},
  year      = {1995},
  pages     = {969-974},
  doi       = {10.1109/ICCV.1995.466829},
  url       = {https://mlanthology.org/iccv/1995/stewart1995iccv-expected/}
}