Presentation 2012/1/12
Hyper-Renormalization
KENICHI KANATANI, ALI AL-SHARADQAH, NIKOLAI CHERNOV, YASUYUKI SUGAYA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We reformulate iterative reweight and renormalization for geometric estimation in computer vision and propose "hyper-renormalization", further improving renormalization. We show the following: Iterative reweight starts from least squares, renormalization starts from the Taubin method, and hyper-renormalization starts from HyperLS. For all, the covariance matrix of the resulting solution achieves the theoretical limit except for high order noise terms. Iterative reweight yields large bias, and renormalization substantially reduces bias, while hyper-renormalization produces no bias achieves the theoretical limit except for high order noise terms and hence is more accurate than maximum likelihood, which is widely regarded as the best method.
Keyword(in Japanese) (See Japanese page)
Keyword(in English)
Paper # Vol.2012-CVIM-180 No.23
Date of Issue

Conference Information
Committee CQ
Conference Date 2012/1/12(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Communication Quality (CQ)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Hyper-Renormalization
Sub Title (in English)
Keyword(1)
1st Author's Name KENICHI KANATANI
1st Author's Affiliation Department of Computer Science, Okayama University()
2nd Author's Name ALI AL-SHARADQAH
2nd Author's Affiliation Department of Mathematics, University of Mississippi
3rd Author's Name NIKOLAI CHERNOV
3rd Author's Affiliation Department of Mathematics, University of Alabama
4th Author's Name YASUYUKI SUGAYA
4th Author's Affiliation Department of Information and Computer Sciences, Toyohashi University of Technology
Date 2012/1/12
Paper # Vol.2012-CVIM-180 No.23
Volume (vol) vol.111
Number (no) 378
Page pp.pp.-
#Pages 8
Date of Issue