Generation of Wide-Range Virtual Environment by Using Photo-realistic Images

Abstract

In this paper, we propose a new approach to generate novel images by using 3-D layer presumption method. In order to extract layers, our proposed method begins with a pair of images with position and orientation information. We zoom and rotate one image in several ways, then compare the changed image with another image. Next, the similar part of regions can be extracted. Lastly, each plane which is texturemapped with the each image of extracted region is arranged to the distance calculated by zooming rate. After the arrangement, we can observe the scene from various viewpoints. The strength of our approach is the ability to avoid correspondence problem inherent in 3-D reconstruction procedure. We think this method is effective to develop wide-range photo-realistic virtual environment by using vehicle-mounted cameras.