Disparity Morphing and Automatic Generation of Stereo Panoramas for Photo-Realistic Virtual Reality Systems

Ho-Chao Huang and Yi-Ping Hung

psfileTR-IIS-97-002


Keywords:
Image Morphing, Image Disparity, Stereo Imaging, Panoramic Imaging, Virtual Reality, Stereo Panorama, Automatic Image Stitching, Epipolar Geometry, Fundamental Matrix

Abstract

Stereo perception is important for a truly immersive virtual reality (VR) system. Recent improvements in computer and video technology have made stereoscopic display systems more prevalent than ever. However, most panoramic imaging systems which can display photo-realistic VR world still have difficulties in automatic production of high-quality stereo panoramas. This paper presents a stereo panoramic imaging system (SPISY) that can automatically generate high-quality photo-realistic stereo panoramas. The major difficulty of generating a stereo panorama from a sequence of stereo images taken by two cameras comes from the image disparity between the two adjacent images to be stitched together, which is caused by the inherent restriction that only one of the two cameras can have its lens center passed by the rotation axis. In this paper, we first analyze the image disparity problem caused by the dislocation of the lens center from the vertical rotation axis, and then propose a disparity morphing algorithm to solve it. Other characteristics of the SPISY include the capability of generating complete-focus views and balanced-exposure views, and of correcting the effect caused by camera tilt and epipolar line inconsistency. With the stereo and automatic correction features, the SPISY can easily provide realistic 360-degree panoramic stereo views for image-based VR systems.