Presentation | 2006-03-17 High Accuracy Fundamental Matrix Computation and Its Performance Evaluation Yasuyuki Sugaya, Kenichi Kanatani, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | This paper studies numerical schemes of maximum likelihood estimation for computing the fundamental matrix from feature point correspondences over two images. First, we state the problem and the associated KCR lower bound. Then, we describe the algorithms of three well-known methods, FNS, HEIV, and renormalization, to which we add a new algorithm based on Gauss-Newton iterations. Using simulated images, we compare their convergence properties. The initial value is chosen in three ways: randomly, by least-squares, and by the Taubin method. We also show real image experiments. These experiments reveal characteristics of each method. It is concluded that FNS has the best convergence properties. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | |
Paper # | PRMU2005-264 |
Date of Issue |
Conference Information | |
Committee | PRMU |
---|---|
Conference Date | 2006/3/10(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 | Pattern Recognition and Media Understanding (PRMU) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | High Accuracy Fundamental Matrix Computation and Its Performance Evaluation |
Sub Title (in English) | |
Keyword(1) | |
1st Author's Name | Yasuyuki Sugaya |
1st Author's Affiliation | Department of Computer Science, Okayama University() |
2nd Author's Name | Kenichi Kanatani |
2nd Author's Affiliation | Department of Computer Science, Okayama University |
Date | 2006-03-17 |
Paper # | PRMU2005-264 |
Volume (vol) | vol.105 |
Number (no) | 674 |
Page | pp.pp.- |
#Pages | 8 |
Date of Issue |