Hybrid rendering and processing for volume visualization

This work involves architectures and algorithms for volume rendering and processing which enhance a scientist's ability to examine volumetric data. Such volume processing as segmentation and mixing with polygonal data provide the scientist with a more powerful paradigm for understanding the data than simply rendering a volume dataset by itself. This extends today's first generation of volume visualization architectures with additional functionality and usability features.; The rendering of perspective projections allows the visualization user to view the dataset from very close range or even from the inside. Current real time systems only allow parallel projections. Two algorithms are introduced, one which utilizes a multipass approach to approximate perspective and another which calculates a high quality accurate perspective sampling by adapting the sampling rate to stay relative to the underlying volume resolution. Both are amenable to implementation on Cube-based architectures.; It is also shown how to connect a volume rendering pipeline with a 3D surface graphics pipeline so that hybrid scenes containing both polygonal and volumetric data can be rendered at interactive to real time frame rates. The method proposes minimal changes to the current architectures while providing an elegant solution to the problem of rendering data of different types. Achieving interactive frame rates with both opaque and translucent polygons empowers the users to analyze 3D spatial relationships among data modelled with different paradigms.; Further, the PAVLOV architecture is presented, which was developed to allow flexible volume processing and rendering within the same accelerator board. It can perform volume segmentation and rendering of the segmented results at interactive to real time frame rates. In this paradigm, volume processing parameters are adjusted in real-time while the user reviews the results of the processing and viewing. Consisting of a 2D mesh of programmable SIMD processing elements, PAVLOV additionally adds flexibility and extensibility into a single complete volume visualization co-processor.; These innovations improve volume visualization architectures to provide functionality beyond merely rendering volume datasets. They allow the scientist to extract more information from data than is currently possible. These innovations allow the user to explore data more efficiently.