Recognition of Shape and Generation of Free Viewpoint Image from Image Database

Haruki Kawanaka¹*, Yuji Iwahori², Nobuaki Sado³ and Kenji Funahashi¹

¹Nagoya Institute of Technology, Gokiso-cho, Showa-ku, Nagoya 466-8555, Japan
²Chubu University, 1200 Matsumoto-cho, Kasugai, Aichi 487-8501, Japan
³TOHO GAS Information System Co., Ltd., 19-18 Sakurada-cho, Atsuta-ku, Nagoya 456-0004, Japan
*E-mail address: haruki@center.nitech.ac.jp

(Received November 30, 2004; Accepted February 10, 2005)

Keywords: Free Viewpoint Image, Silhouette, Principal Component Analysis, Pose Recognition

Abstract. For the purpose of generating a free viewpoint image in the scene, the previous studies need many actual cameras in the stadium. These studies take large scale environment with much cost, including actual camera environment such as a motion capture system. In this paper, a low cost and general approach is developed to generate a virtual image at another viewpoint from multiple viewpoint image database. This multiple viewpoint image database is designed for a soccer player and used to generate a virtual scene.