Generation of the 3-dimensional picture-pair which follow the motion of observer

Abstract

This paper presents the generation of an azimuth difference picture pair for a 3-dimensional display. It is with a set of two video cameras, and parallax-images captured by the each camera. From the correspondence relation between parallax-images pair, each pixel depth is computed and the depth map is created. A 3-dimensional model is created from the depth map, and parallax-images generated from a different eye point. An observer's eye position is measured on real time by the pupil position detection method developed at our laboratory. By changing parallax-images in accordance with a motion of an observer, the 3-dimensional display with high realistic sensation is enabled.