Presentation 2002/10/11
Probabilistic Relaxation Method in Multi-Resolusion Space for Regularization of Correspondence Detection Problem
Jun MATSUO, Hiroyuki KUDO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Correspondence detection problem is an inverse problem which needs a regularization method. Untill now, this problem was solved by using the local smoothing constraint. But, unfeatured regions cannot be smoothed sufficiently and edges are over-smoothed in these methods. In this paper, we propose a correspondence detection method which is based on propagating solutions in multi-resolution space. This method can control degree of smoothing in unfeatured region and edges. Low resolution solution includes false correspondence, but this problem is solved by applying the probabilistic relaxation in multi-resolution.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Correspondence detection problem / regularization method / multi-resolution space / probabilistic relaxation
Paper # NC2002-56
Date of Issue

Conference Information
Committee NC
Conference Date 2002/10/11(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 Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Probabilistic Relaxation Method in Multi-Resolusion Space for Regularization of Correspondence Detection Problem
Sub Title (in English)
Keyword(1) Correspondence detection problem
Keyword(2) regularization method
Keyword(3) multi-resolution space
Keyword(4) probabilistic relaxation
1st Author's Name Jun MATSUO
1st Author's Affiliation Doctoral Program in System Information Engineering, University of Tsukuba()
2nd Author's Name Hiroyuki KUDO
2nd Author's Affiliation Institute of Information Sciences and Electronics, University of Tsukuba
Date 2002/10/11
Paper # NC2002-56
Volume (vol) vol.102
Number (no) 382
Page pp.pp.-
#Pages 6
Date of Issue