Klaus Mueller
Professor

Director, Visual Analytics and Imaging (VAI) Lab
Liaison, SUNY Korea CS Program

Center for Visual Computing
Computer Science Department
Stony Brook University - State University of New York

mueller{remove_this}@cs.stonybrook.edu

Visual aids for high performance code developers: PUMA-V: Optimizing Parallel Code Performance Through Interactive Visualization soon in IEEE CG&A

Medical imaging. Here we use programmable commodity graphics hardware boards (GPUs) to accelerate a wide variety of 3D computer tomographic (CT) reconstruction algorithms. So far we have achieved speed-ups of 1-2 orders of magnitude, without significant loss in reconstruction quality. Our RapidCT system enables fast, accelerated 3D reconstruction in diagnostic imaging, radio-therapy applications, surgery planning, electron microscopy, and others, at a fraction of the cost of proprietary devices.

Related projects are the iterative 3D reconstruction from data acquired with X-ray CT scanners at low radiation doses (known as low-dose CT), with transmission ultrasound for breast mammography, data obtained from MV-CT and proton-CT scanners for the treatment of cancer, projections obtained via mobile X-ray source/detector pairs, as well as functional imaging applications, including MRI, functional MRI, SPECT, and PET.

Scientific, medical, and information visualization. This area embraces a wide gamut of projects. In volume visualization, we are concerned with algorithms and techniques for volume rendering (point-based and ray-based) on regular as well as irregular grids, fundamental research on interpolation filters and data grids, GPU accelerated rendering, rendering of multi-modal and time-varying datasets, feature-centric and illustrative visualization, intuitive user interfaces, modeling with volumetric datasets (examples: ablation, melting), volume graphics and rendering with volumetric effects (examples: radiosity, shadows), image-based volume rendering, and others. In information visualization, we are working on the development of frameworks for the visualization of large, high-dimensional, multi-valued datasets.

Modeling of natural phenomena. We have developed a comprehensive framework for the modeling and simulation of smoke, fire, and general gaseous phenomena, both on surfaces and in 3D space, interacting with static and moving objects. Our approach uses the Lattice-Boltzmann Model (LBM) for fast propagation of these phenomena. The LBM is non-iterative and uses only local operations per grid update of the transient phenomena. It is therefore very attractive for acceleration on GPUs. Current applications are the modeling of gaseous substances in urban (homeland) security scenarios, the simulation of complex heat-originated phenomena for computer graphics, such as heat shimmering, melting, ablation, and smoke, and the simulation of other amorphous phenomena.

General purpose computing on programmable graphics hardware (GPGPU). This is a topic related to the above, but not confined to it. We have been using GPUs for various costly numerical simulation tasks (in addition to our GPU acceleration of medical imaging algorithms), with speedups of 1-2 orders of magnitudes, while being able to use the GPU also to quickly visualize the (intermediate) results. This gives rise to the notion of visual simulation.