Sadagić, Amela; (1998) Efficient Image Display for Head-Slaved Viewing of Virtual Environments. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Efficient_image_display_for_he.pdf Download (20MB) | Preview

Abstract

The efficient resolution of visibility in polygonal scenes was and still is one of the most demanding tasks in computer graphics. A need for faster algorithms is becoming even more important to fulfil the requirements of real-time Virtual Reality (VR) systems, and in particular immersive VR systems. In VR systems, typically the viewpoint is changing at a greater rate than the objects within the scene. In such environments the visibility, in terms of back-to-front polygon visibility ordering, does not change dramatically from frame to frame since the viewpoint changes by small increments. In addition, stereopsis requires visibility determination for each eye; again, since the viewpoints corresponding to the eyes are so close together, coherence can be exploited to avoid the necessity of separate visibility computations for each eye. These considerations provide a sound basis from which to investigate the areas where efficiency of visibility computation might be improved. This thesis therefore considers a particular class of visibility algorithm that pre-order the scene in visibility order with respect to the viewpoint. The thesis investigates the construction of visibility algorithms that makes use of the resulting viewpoint movement coherence, where visibility information computed with respect to one viewpoint may be incrementally adjusted for a subsequent viewpoint without the necessity for a full visibility computation for each frame. An algorithm has been designed which supports space partitioning into cells with the property that each cell has an implicitly associated invariant visibility ordering. We introduced the concept of dynamic priority ordering. It consists of the initial ordering, generated as a linear list in the preprocessing phase, and a set of rules which define dynamic updates of the priority ordering to the one that is valid for the current viewpoint position. The entire space within which the viewpoint moves is called Viewpoint Movement Space (VpMS). This may be parameterised - it can be defined as a subset of ID (line segment or curve), 2D (convex polygon) or 3D space (convex polyhedron). A data structure which embeds both geometrical as well as topological information has been devised. It supports underlying spatial subdivision and provides very fast retrieval of information necessary to update the visibility list. Methods for utilising the same data structure for collision detection and image based rendering are discussed. Finally, experimental results, where the VpMS algorithm has been applied on a range of scenes, showed improvement in the frame rate when compared to the use of BSP trees.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as JSONCSVXML

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item