Research And Implementation Of Image-Based Virtual Space

Free Image-Based Virtual Space (FIBVS) is a kind of virtual space, which supports free walkthrough by organizing real world images or videos captured by cameras.Compared with traditional geometry-based virtual space, image-based virtual space (IBVS) has some advantages: constructing easily, rendering fast and reality, et al. However, the free degree of IBVS is limited by real images. How to walk freely in IBVS is a key issue of the research of IBVS construction, so this thesis puts moreconcerns on it. To build an ideal IBVS system , we must resolve the following main issues:(1) Establishing the architecture of theory and technique for IBVS.(2) Modeling approaches of IBVS.(3) 3D spatial operation approaches.(4) The fusement of temporal property and spatial property.(5) The compound of multiple data source of different resolution, different focus and different time.The main contents and contributions of the dissertation include:(1) We put forward a kind of theory and some techniques for FIBVS to achieve free walkthrough. We study the characteristics, types, components, space organization and manipulation of FIBVS, and solve the problems of free walkthrough, manipulation of image-based virtual objects, and extraction of all kinds of multimedia information.(2) We put forward a model of information organization and operation of IBVS. The building of virtual space refers to the mode of hypermedia which maker use of the structure of node and hyperlink to organize local images. Comparing with establishing spatial model by polygon, this structure much more fits for the ready-made images and the realistic virtual environment. Furthermore, by taking the temporal property of spatial presentation, we discuss the fuse issues of temporal and spatial properties of IBVS.(3) We modify and implement the algorithm of producing 360-degree cylindrical panoramic image. We improves on some local algorithms and integrates our own algorithms toform a new set of algorithms for creating panoramic image, which include cylindrical projection algorithm, image matching algorithm, seamless linking algorithm and cylindrical to planer projection algorithm.