HyperLS for parameter estimation in geometric fitting

Kenichi Kanatani, Prasanna Rangarajan, Yasuyuki Sugaya, Hirotaka Niitsuma

Research output: Contribution to journalArticle

14 Citations (Scopus)

Abstract

We present a general framework of a special type of least squares (LS) estimator, which we call "HyperLS," for parameter estimation that frequently arises in computer vision applications. It minimizes the algebraic distance under a special scale normalization, which is derived by a detailed error analysis in such a way that statistical bias is removed up to second order noise terms. We discuss in detail many theoretical issues involved in its derivation. By numerical experiments, we show that HyperLS is far superior to the standard LS and comparable in accuracy to maximum likelihood (ML), which is known to produce highly accurate results but may fail to converge if poorly initialized. We conclude that HyperLS is a perfect candidate for ML initialization.

Original languageEnglish
Pages (from-to)80-94
Number of pages15
JournalIPSJ Transactions on Computer Vision and Applications
Volume3
DOIs
Publication statusPublished - 2011

Fingerprint

Parameter estimation
Maximum likelihood
Error analysis
Computer vision
Experiments

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition

Cite this

HyperLS for parameter estimation in geometric fitting. / Kanatani, Kenichi; Rangarajan, Prasanna; Sugaya, Yasuyuki; Niitsuma, Hirotaka.

In: IPSJ Transactions on Computer Vision and Applications, Vol. 3, 2011, p. 80-94.

Research output: Contribution to journalArticle

Kanatani, Kenichi ; Rangarajan, Prasanna ; Sugaya, Yasuyuki ; Niitsuma, Hirotaka. / HyperLS for parameter estimation in geometric fitting. In: IPSJ Transactions on Computer Vision and Applications. 2011 ; Vol. 3. pp. 80-94.
@article{9cdcbd94eafd41cf800997cdf119a1f9,
title = "HyperLS for parameter estimation in geometric fitting",
abstract = "We present a general framework of a special type of least squares (LS) estimator, which we call {"}HyperLS,{"} for parameter estimation that frequently arises in computer vision applications. It minimizes the algebraic distance under a special scale normalization, which is derived by a detailed error analysis in such a way that statistical bias is removed up to second order noise terms. We discuss in detail many theoretical issues involved in its derivation. By numerical experiments, we show that HyperLS is far superior to the standard LS and comparable in accuracy to maximum likelihood (ML), which is known to produce highly accurate results but may fail to converge if poorly initialized. We conclude that HyperLS is a perfect candidate for ML initialization.",
author = "Kenichi Kanatani and Prasanna Rangarajan and Yasuyuki Sugaya and Hirotaka Niitsuma",
year = "2011",
doi = "10.2197/ipsjtcva.3.80",
language = "English",
volume = "3",
pages = "80--94",
journal = "IPSJ Transactions on Computer Vision and Applications",
issn = "1882-6695",
publisher = "Information Processing Society of Japan",

}

TY - JOUR

T1 - HyperLS for parameter estimation in geometric fitting

AU - Kanatani, Kenichi

AU - Rangarajan, Prasanna

AU - Sugaya, Yasuyuki

AU - Niitsuma, Hirotaka

PY - 2011

Y1 - 2011

N2 - We present a general framework of a special type of least squares (LS) estimator, which we call "HyperLS," for parameter estimation that frequently arises in computer vision applications. It minimizes the algebraic distance under a special scale normalization, which is derived by a detailed error analysis in such a way that statistical bias is removed up to second order noise terms. We discuss in detail many theoretical issues involved in its derivation. By numerical experiments, we show that HyperLS is far superior to the standard LS and comparable in accuracy to maximum likelihood (ML), which is known to produce highly accurate results but may fail to converge if poorly initialized. We conclude that HyperLS is a perfect candidate for ML initialization.

AB - We present a general framework of a special type of least squares (LS) estimator, which we call "HyperLS," for parameter estimation that frequently arises in computer vision applications. It minimizes the algebraic distance under a special scale normalization, which is derived by a detailed error analysis in such a way that statistical bias is removed up to second order noise terms. We discuss in detail many theoretical issues involved in its derivation. By numerical experiments, we show that HyperLS is far superior to the standard LS and comparable in accuracy to maximum likelihood (ML), which is known to produce highly accurate results but may fail to converge if poorly initialized. We conclude that HyperLS is a perfect candidate for ML initialization.

UR - http://www.scopus.com/inward/record.url?scp=84855415759&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84855415759&partnerID=8YFLogxK

U2 - 10.2197/ipsjtcva.3.80

DO - 10.2197/ipsjtcva.3.80

M3 - Article

VL - 3

SP - 80

EP - 94

JO - IPSJ Transactions on Computer Vision and Applications

JF - IPSJ Transactions on Computer Vision and Applications

SN - 1882-6695

ER -