Neophytos Neophytou - Research

I currently work with prof. Klaus Mueller at the Center for Visual Computing at the Computer Science Department at Stony Brook University, and prof. Wei Zhu, who is leading the Proteomics Research group at the department of Applied Math and Statistics. I also participate in some collaborative projects with the Computer Science department at the University of Cyprus with prof. Paraskevas Evripidu.

Following is a list of most of the projects that I have been working on. A complete list of publications can be found in my Publications page.

Dissertation:

A Generalized Framework for Interactive Volumetric Point-Based Rendering. We present our efforts on Point-Based Volume rendering and specifically on the image-aligned post-shaded splatting algorithm which was proposed as a remedy to the drawbacks of existing algorithms and gave special focus on image quality. In the course of this dissertation, we follow the evolution of this algorithm through several stages of maturity.

GPU-accelerated Image Processing Tools:

	Colorspace CAD. We introduce a GPU accelerated framework for image color transformations based on interactive manipulation of 3D gamuts in the colorspace domain.

Real-time Rendering:

	SIMD-Aware Ray-casting. We study the latest single-pass ray-casting algorithm with specific attntion to the characteristics of the underlying SIMD GPU architecture. Our proposed solution achieves speedups of up to 8-10 times.

Splatting and GPU-Accelerated Splatting:

	Rendering of regular volumes using GPU accelerated splatting. We have ported the image aligned splatting algorithm to a GPU-aware implementation. We enable empty space skipping and early splat elimination by employing an elaborate mechanism to take advantage of GPU specific extensions that use early-z-culling to reduce extraneous fragment processing.
	Rendering of irregular volumes using GPU accelerated splatting. We generalize our previous GPU image aligned splatting implementation for irregular volumetric datasets. We generalize the uniform spherical kernel primitive to an arbitrary ellipsoidal kernel and provide a mechanism for the slicing and rendering of ellipsoid-based irregular datasets on the GPU.
	Post Convolved Splatting. We use post-convolution to accelerate rendering at high magnifications for both X-Ray and Post-Shaded image-aligned Splatting.
	Space-Time points: 4D Splatting using efficient grids. We use 4D-BCC grids to achieve up to 50% lossless space compression and up to 30% of effective rendering time reduction for time-resolved volumetric datasets.

Volume Rendering Techniques:

	SQUEEZE: Numerical-precision-optimized volume rendering. We investigate how to squeeze volume rendering into as few bits per operation as possible while still retaining excellent image quality. Both the raycasting and splatting volume rendering pipelines are discussed.

With the Proteomics Research Group at the
Department of Applied Math and Statistics:

	proteoExplorer is a complete environment for the analysis and visualization of large-scale proteomic mass spectrum datasets. The visualization module employs a multi-resolution approach that allows the interactive display of up to hundreds of very high resolution mass-spec datasets in a very innovative user environment.
	The ViStaMS: Visual Statistical Data Analysis for Microarray and SAGE data provides an easy to use and accessible environment for the analysis of large-scale genetics microarray and SAGE datasets. It is implemented as a MS-Excel addin and it is currently being used by the Osteosarcoma Research Laboratory at the Montefiore Children's Hospital, Bronx - New York.

With the Parallel Architectures group
at the University of Cyprus:

	The Netdbx parallel debugger is a Java-powered tool for debugging MPI-based parallel programs accross low-bandwidth internet lines. It provides a layered multi-platform architecture which allows the user to run the program within an internet browser.
	Netdbx-G extends the Net-dbx framework to enable the debugging of Grid-based parallel applications. The system architecture has been extended to interface with grid services and provide a direct link to the running nodes to provide real-time source level debugging.
	MPI-FT is one of the earliest attempts to implement a fault tollerant framework for the MPI library for Parallel/Distributed computation. This prototype framework is currently outdated. It is referenced, however, as one of the first implementations to even deal with the Fualt tollerance issues of MPI.

---

Go back to Home Page, Publications