Efficient and Comprehensible Visualization of Complex 3-D Scenes (thesis)
Abstract:
The advancement of computer technology has brought about greater ability to model and render complex 3-D scenes. Highly complex structures in these scenes give rise to large amounts of information that can be hidden using conventional rendering and visibility techniques because of occlusion. In order to expose structural information in a complex model, non-photorealistic visibility techniques must be used.
This thesis presents non-deformative techniques for creating structurally rich visualizations of complex 3-D scenes. Techniques for reducing occlusion globally are combined with an importance specification system to expose and emphasize objects and regions of interest. These occlusion reduction methods are also applied locally according to an occlusion function defined over a cutaway structure. Localized occlusion reduction can be combined with augmentations to the polygonal and volumetric rendering pipeline to create cutaway illustrations that expose and emphasize objects and regions of interest. The presented techniques are made with realtime rendering in mind, and are shown to operate at interactive framerates for dynamic models and scenes. Sample result images are presented for medical and architectural visualization scenarios.
The main contributions of this thesis are a novel method for creating sparse line representations of isosurfaces in volumes, a contextual cutaway structure and associated occlusion function, an efficient and flexible image-based cutaway surface computation algorithm, and several cutaway surface function modifications to support perspective projection correction, edge compression, locally defined angles, and directional constraints.