The Multi-LDI: an image based rendering approach for interaction navigation, and visualization in complex virtual environments

Abstract

In this paper, we present a new data structure for image-based rendering: the multi LDI. The multi LDI consists of a number of Layered Depth Images (LDI) covering a hemisphere of possible viewing angles. It allows compact image-based storage and fast rendering of large and complex scenes, while supporting rendering from large range of viewing directions. Since the internal occlusion in each LDI is very small, and only one of the LDIs in the multi LDI is rendered at a time, the rendering cost is significantly reduced compared to geometry-based rendering and even compared to other image-based rendering methods. Moreover, the single LDIs in the multi LDI can be generated on demand using a number of depth images rendered with an offscreen renderer. We also discuss a comparison of the geometry-based rendering and our image-based method and present some measured rendering times. Furthermore, we describe the utilization of this technique in a complex Virtual Environment (VE) for realizing navigation, visualization, and interaction aids. In particular, we present a multiple viewport technique providing two important features: (1) A sort of history during the modeling process, whereas a live 3D copy of the scene is displayed in a window in front of the user. (2) Different live views of the scene, seen from arbitrary viewpoints, in order to display details occluded in the normal view.