View synthesis by assembling parts of pre-acquired images

作     者：Katsuyuki Kamei Minoru Maruyama Kazuo Seo 

作者机构：Advanced Technology R&D Center Mitsubishi Electric Corporation Amagasaki Japan Faculty of Engineering Shinshu University Nagano Japan 

出 版 物：《Systems and Computers in Japan》 

年 卷 期：1999年第29卷第9期

主　　题：virtual environment view synthesis image-based rendering image stripe Peano scan 

摘      要：This paper presents a newly developed image-based rendering technique for constructing a virtual environment. The method synthesizes a new view by splitting source images and assembling the divided sections (stripes) according to the viewing rays. The most important feature of the method is that it does not need a precisely-captured large set of source images. Instead, it focuses on the object and synthesizes a new image by selecting the most suitable stripe among the pre-acquired source images. This makes it possible to use only a limited set of source images. Also, it eliminates preliminary processes, such as obtaining correspondences between multiple images. As a result, it makes it easier to input a natural scene into the computer and to reproduce it as a virtual environment. We also describe improvements in image quality obtained by smoothing, examination of errors in the synthesized image, and a technique for arranging input points of source images. Moreover, an efficient method for data management is investigated. Fine images can be synthesized for a viewpoint that moves in any horizontal direction, so that a virtual environment with powerful walkthrough operations can be easily constructed. © 1998 Scripta Technica, Syst Comp Jpn, 29(9): 1–12, 1998