During recent years, the use of virtual reality technology has become widespread and popular. However, to further broaden the application of virtual reality, more sophisticated and realistic virtual worlds need to be developed. Traditionally, most virtual worlds are generated using three-dimensional (3D) computer graphics incorporating 3D geometric models and various rendering software. However, if 3D models become very complex, the delay time caused by rendering calculations makes it difficult for the user to be able to interact with the virtual world. Also, the production of realistic 3D computer graphics is very cost and labor intensive. From a very practical point of view, it is clear that we need some alternate approaches to realize a truly realistic virtual world. In this paper, the authors introduce an alternate method of generating virtual worlds other than 3D computer graphics. The method discussed here is to generate virtual worlds by processing 2D real images taken by video cameras. For this purpose, a special video camera system that can record image data indexed by position data was developed. Using recorded image data indexed by position data we are able to experience the virtual image world interactively. This method has become realistic due to advances in multimedia computers capable of handling large image data. A tested prototype of this kind of system is discussed in some depth, along with the capability and limitations of this prototype system.