Doctor of Philosophy (PhD), Ohio University, 2013, Electrical Engineering & Computer Science (Engineering and Technology)

Humans live in a 3D reality and experience a 3D world. As a result, 3D computer graphics is a natural way to present digital data to humans. Virtual worlds of great complexity are naturally perceived and understood by the human visual system. This mechanism offers an efficient pathway to transfer digital data from a display into human knowledge. Computing 3D rendering output requires a specific computational complexity for a specific scene. The ever increasing demand to expand a scene's boundaries, add additional details, and enhance a scene's behavior present challenges to both the rendering hardware and rendering algorithms. This work presents a set of novel image-based algorithmic rendering approaches designed for massive concurrent execution on modern programmable Graphical Processor Units (GPUs). These approaches make use of constructs known as billboards and mesostructures, their goal is to render as much interactive detail as possible while maintaining real-time framerates. First, billboards are used to dynamically generate image-based impostors of computationally expensive objects within a scene. These impostors are subsequently rendered in lieu of the original geometry thus reducing the scene's overall rendering time. Impostors are then applied to the real-time visualization of tens of millions of 3D Light Detection and Ranging (LIDAR) data points enabling highly interactive visualizations previously unable to be achieved on commodity hardware. Second, mesostructures are used to dynamically generate a 3D height field on top of an existing geometrical surface via a 2D texture known as a displacement map. This work presents a technique for interactively deforming and colliding with mesostructures at a per-texel level. This technique integrates well with existing physics engines and is able to reduce traditional 3D geometrical deformations (vertex-based) to 2D image space operations (pixel-based) that are parallelized on a GPU without CPU-GPU data shuffl (open full item for complete abstract)

Committee: David Chelberg PhD (Advisor) Subjects: Computer Science