From Transfer Functions to Level Sets
Advanced Topics in Volume Image Processing

IEEE Visualization 2001

Tutorial 5

Contents

This is a directory for the supplemental materials for the IEEE Visualization 2001 tutorial on From Transfer Functions to Level Sets: Advanced Topics in Volume Image Processing. These materials were prepared for the Vis2001 DVD-ROM proceedings. Additional information and any materials added after production of the DVD-ROM can be found in the current online repository for the course at http://visual.nlm.nih.gov/tutorials/vis2001

Abstract

The conventional volume rendering pipeline has been effectively used to visualize volume data that is often considered a sampled density map. However, more and more people are looking at data that has noise, occluding surfaces, density fluctuations, limited resolution, etc. These factors require users to do more "processing". Advanced volume processing is what enables people to do 1) linear and nonlinear filtering, 2) interpolation, 3) reconstruction, 4) feature extraction, and 5) model fitting. We describe the problem as a pipeline from the reconstruction of the continuous model from the sampled data, through the application of transfer functions for shading and classification, to the transformation sampling and projection of the reconstructed values for visualization. The goal is to extract or locate structures hidden within the data. A tacit requirement is to do so without masking detail with unwanted artifacts. Thus, the emphasis will be on factors which affect final image quality.

Beyond improving the volume visualizations that we are used to seeing, faster systems are allowing users the freedom to explore and interact with their data. Designers of visualization systems are supplementing viewpoint and clip-plane control with a variety of interactive tools for controlling opacity, color, texture, and other attributes of the presented image. Moreover, more sophisticated means of analyzing volume data leads to broader dimensions of the visualization space. Complex data requires more than simple isosurfaces, and effective visualization requires a blend of mathematics, statistics, and aesthetic design to quickly and clearly convey the intended message. We are proposing a course that will cover elements all along the volume rendering pipeline. Beyond an introduction to volume rendering, we will target specific problems encountered in the creation of volume visualizations and the mathematics required to address them. This is not intended as a superficial survey course on volume mathematics, but rather a series of studies designed to take the attendee through many of the deep problems in volume visualization. Case studies and examples are an integral part of the course. We will also present methods for navigating and interactively exploring volume data through the use of transfer functions, level sets and implicit models.

Emerging topics in volume mathematics

The afternoon will be dedicated to exploring emerging techniques relevant to both 3D image processing and volume graphics. Proposed topics for discussion include multiscale methods, implicit techniques, and level set theory. These methods represent active areas of research that should interest audience members who wish to explore new ideas in volume graphics and visualization research. For instance, multiscale methods, will be described as a natural extension of the filtering techniques for function reconstruction described earlier in the morning. The advantages of these techniques will be illustrated for analysis through suitable examples. Wavelet techniques will be described. Less emphasis will be paid to the actual design issues of such filters. Rather, the emphasis will be on the utility of these techniques.

Index to Course Materials

Printed Course Materials, Handouts

Supplemental Course Materials

Papers on Reconstruction, Transfer Functions, and Parameter Modeling

Papers on Wavelets and Multiscale Analysis

Index of Online Resources

Online Tutorials and Materials (list under development)

Indices to online conference pages, journals, and bibliographies

Speaker Information

Speaker Biosketches

Gordon Kindlmann Gordon Kindlmann has been a doctoral student in the Computer Science department at the University of Utah since 1997. In 1995 he received a BA in mathematics from Cornell University, and in 1999 he finished his MS in computer graphics under Donald Greenberg in the Program of Computer Graphics at Cornell University. His Masters research was on the Semi-Automatic Generation of Transfer Functions for Direct Volume Rendering. His current research continues to focus on volume rendering while extending into the areas of medical imaging, segmentation, and color science.

Raghu Machiraju is an Assistant Professor of Computer and Information Science at the Ohio State University. His research interests include visualization, graphics, image analysis and high performance computing. He obtained his Doctorate from the Ohio State University in 1996. Previously he served as an Assistant Professor of Computer Science at Mississippi State and as a research scientist at the NSF ERC for Computational Field Simulation. He received the NSF Faculty Early Career Award for his proposal, "On the Assessment of Volume Rendering Algorithms in Visual Computing" in 1998.

Torsten Möller received his PhD in Computer and Information Science from Ohio State University in June 1999. He received a Vordiplom (BSc) in mathematical computer science from Humboldt University of Berlin, Germany. He is currently an assistant professor at the School Of Computing Science at Simon Fraser University, where he is co-director of the Graphics, Usability and Visualization Lab. His research interests include the fields of Scientific Visualization and Computer Graphics. He is especially interested in interactive and accurate volume rendering methods for regular and irregular data.

Ross Whitaker is currently an assistant professor at the University of Utah, Department of Computer Science. His research interests include: computer vision, image processing, medical imaging, and computer graphics/visualization. He received his B.S. degree in Electrical Engineering and Engineering Physics from Princeton University in 1986 and his Ph.D. in Computer Science from the University of North Carolina at Chapel Hill in 1993. Previously he has been a research scientist in the User Interaction and Visualization Group at the European Computer-Industry Research Centre in Munich, Germany and an assistant professor of Electrical and Computer Engineering at the University of Tennessee.

Terry S. Yoo is a Computer Scientist in the Office of High Performance Computing and Communications, National Library of Medicine, NIH, where he explores the processing and visualizing of 3D medical data, interactive 3D graphics, and computational geometry. Previously as a professor of Radiology, he managed a research program in Interventional MRI with the University of Mississippi. Terry holds an A.B. in Biology from Harvard, and a M.S. and Ph.D. in Computer Science from UNC Chapel Hill.