Generation of True 3D Films

We de ne a true 3D lm as a lm that can be viewed from any point in space. In order to generate true 3D lms, the 3D-MATIC Research Laboratory of the University of Glasgow has been developing a capture 3D studio based on photogrammetry technology. The idea is simply to generate 25 photo-realistic 3D models of a scene per second of lm. After the presentation of the state of the art in the domain, the core technology of our dynamic 3D scanner is detailed. Finally rst results, based on a 12 camera system, are shown and the potential applications of this new technology for virtual storytelling are investigated. 1 What is a true 3D lm? From the 30s onwards cinema has tried to create the illusion of the 3D image by arti cially reproducing binocular vision. The cinema aims to lm separately the same object from two angles corresponding to our binocular vision and to project them onto the same screen. It only remains for each eye to select the image, it is meant to receive, in order to recreate a 3D illusion. Special glasses are used for this purpose. This technology has been quite successful considering that nowadays such 3D lms are o ered every day in cinemas such as IMAX . Since these lms are lmed from a speci c viewpoint, spectators can only see 3D images from that viewpoint. For some people, it may be a bit frustrating, they would like to see these 3D images from other view points: they would like to see true 3D lm. We de ne a "true" 3D lm as a lm that can be viewed from any point in space. Ideally spectators should have the ability to choose interactively the position of the camera while watching these lms; they should be able to y over scenes of a lm, as it is now possible to navigate through 3D virtual reality environment using VRML browsers. In order to generate these true 3D lms, the 3D-MATIC Research Laboratory of the University of Glasgow has been developing a capture 3D studio based on photogrammetry technology. In this paper, we present the prototype of the 3D studio we are currently developing. First we relate our research to previous work in the elds of 3D capture, modelling and animation of virtual humans. Then we describe the technology we developed and nally we show some experimental results and o er some applications for virtual storytelling.