Distortion-free fusion of multiple video camera images using EPI analysis

This paper presents a method that uses an Epipolar Plane Image (EPI) to calibrate multiple video cameras and obtain distortion-free fusion of these images. Generally speaking, the optical center of a camera exists inside its body making it physically difficult to focus these images at one point in multiple cameras. When fusing these types of camera images, distortion will occur unless three-dimensional shape information is used. In this paper we perform distortion-free image fusion by bringing these optical centers into agreement in a spatiotemporal field based on conditions in which multiple video cameras are installed on a movable object. The agreement points of the optical centers are automatically calculated by means of EPI matching. Because this process utilizes time-variant redundancy of each camera image, a stable corresponding relationship is obtained even when there is no spatial overlap of the photographic range making it possible to install cameras with comparative freedom. In addition, because it is possible to ensure sufficient quality without using special external synchronization devices or position measurement devices, this method is also suitable for acquiring textures of large-scale real scenes, such as urban spaces.