“Unified views of motion and illumination in computer vision” (preliminary)

日時:
2008年3月13日 12:30-14:00

場所:
電気系総合館　中会議室
※　各自弁当を持ち込むランチョンセミナーの形式です．
※　以下，講演は日本語で行われる予定です．

講師:
Atsuto Maki氏

講師所属:
Department of Intelligence Science and Technology

講師略歴:


講演概要:
Illumination variance is observed on the surface of an object as long as the object is in motion. I have been investigating the possibility of utilizing such variance for different issues including surface reconstruction, motion segmentation, and constructing linear image basis. In this talk we first introduce a novel method for synthesizing images of a 3d object in arbitrary poses illuminated from arbitrary directions, given a few images of the object in unknown motion. Since an appropriate re-alignment of image pixels is necessary, we solve the correspondence between input images by the Geotensity constraint that accurately governs the relationship between four or more images. We also address the issue of separating the motion of multiple moving objects in an image sequence while proposing the notion of photometric subspace. We show that intensities of observed trajectories of image features on a single body lie on a linearly independent frame space with three, or fewer, dimensions. We then argue that it is possible in theory to determine the grouping of the feature points by way of separating the photometric subspaces. Experimental results are presented for both examples.

備考:
連絡先
グローバルCOEプログラム
「知識循環社会のための情報学教育研究拠点」
原初知識モデル教育研究コア事務担当　永藤美奈子
電話 /FAX : 075-753-4961　　
E-mail : nagafuji於ii.ist.i.kyoto-u.ac.jp　（於を@に置き換えてください）