Per Christian Hansen教授による講演会(その3)
- 投稿者
- 相原 研輔(東京都市大学)
- 日程
- 2020年 2月26日(水) 2:30PM-4:30PM
- 会場
- 東京都市大学(世田谷キャンパス) 6号館1階 61A教室 (アクセス)
| 本文 |
この度，デンマーク工科大学のPer Christian Hansen教授をお招きし，東京都市大学にて下記の講演会を開催することとなりました． ご興味のある方は是非ご参加ください． 本件は東京で開催する3つの講演会の3つ目です．他2件は2月25日に一橋大学にて開催されます． This lecture is supported by Invitational Fellowships for Research in Japan (Short-term S). |
| Speaker |
Professor Per Christian Hansen (DTU Compute, Technical University of Denmark) http://www2.compute.dtu.dk/~pcha/ |
| Title | Algebraic Iterative Reconstruction Methods for X-ray CT and Their Convergence Properties |
| Abstract |
We use algebraic iterative reconstruction methods - such as ART (Kaczmarz), SART, and SIRT - to solve discretized inverse problems in computed tomography.
They are very flexible because the underlying system A x = b requires no assumption about the scanning geometry, and it is easy to incorporate convex constraints
(e.g., box constraints). Their success in computing regularized solutions is due to a mechanism called semi-convergence.
When we implement these methods with a focus on computational efficiency, we often use different discretization schemes for the forward projection and the back projection.
This means that there is a mismatch between the back projection matrix B and the transposed of the forward projection matrix A. The use of such an unmatched A,B-pair has two consequences:
the accuracy deteriorates (compared to when using a matched pair), and the iteration may fail to converge.
In this talk, I will survey some recent results related to the convergence and semi-convergence of the algebraic iterative reconstruction methods. I will also present a novel approach to "fixing" the non-convergence issue with only a small computational overhead.
I illustrate the theory with numerical results. This is joint work with Tommy Elfving, Touraj Nikazad, Yiqiu Dong and Nicolai A. B. Riis. |
お問い合わせ先
阿部 邦美，相原 研輔
e-mail: abe@gifu.shotoku.ac.jp , aiharak__AT__tcu.ac.jp