Dr. Yarden Livnat

Research Assistant Professor, School of Computing
Adjunct Assistant Professor, School of Medicine
Scientific Computing and Imaging Institute
WEB 4652
SCI Personnel Page

Curriculum Vitae: cv.pdf


Information Visualization Scientific Visualization

Epinome: A Visual-Analytics Workbnech for Epidemiology Data

Intuitive Visual Representation of Network Events to Improve Decision Making:

The goal of this research is to develop a new generation of cybersecurity visual displays that integrate distinct and advance in data representation design, cognitive psychology, computer visualization, with heuristic knowledge and statistical methods from cybersecurity experts, to significantly improve network analysts' and decision makers' ability to discover, diagnoses and rapidly respond to critical cyber situations.

Scientific Portals - Remote Visualization:

Isosurface extraction is an important visualization tool for understanding underlying structures in volumetric data sets. The use of supercomputers enable scientists to perform simulations on very large data sets, however the enormous data size overwhelm the currently available interactive isosurface extraction and rendering applications. The recent availability of inexpensive yet powerful desktop computers and parallel supercomputers in limited locations suggests that utilization of these resources via a new visualization paradigm namely, remote isosurface extraction, will allow interactive visualization of large data sets.

In this work, we present a new framework in which a scientist can access sophisticated visualization methods instaniated on a remote supercomputer through a portal. This portal provides the scientist, sitting at his desktop, access to a view-dependent volume visualization tool enabling a dynamic visualization of only the visible portion of an isosurface . The local nature of the portal enables interactive visualization while the portal back-end negotiates with the remote supercomputer to augment the local portion of the isosurface base on the changing view point of the user.

Pisa: Point Based Isosurface Algorithm:

In this paper [pdf] we propose a novel point-based approach to view dependent isosurface extraction. We also introduce a fast visibility query system for the view dependent traversal, which exhibits moderate memory requirements. Using this approach we achieved an interactive interrogation of the full visible woman dataset (1GB) at more then 4 frames per seconds on a desktop computer.

The point based approach is based on an extraction scheme that classify different sections of the isosurface into four different categories. The classification is based on the size of the geometry when projected onto the screen. In particular, we use points to represent small and sub-pixel triangles as well as large sections of the isosurface which project to less then a pixel on the screen.

An important issue in point based representation is how to define a normal to a point that represent a large, but far, sections of the isosurface. We propose to define such normals based on post processing of the extracted isosurface and show a hardware implementation using a vertex and fragment programs. The ability to define the normals using screen space connectivity opens the door to further image space effects such as boundary enhancements that are especially vital for complex isosurfaces.

Shadows for Incomplete Point-based Isosurfaces:

This paper [pdf] presents a method for computing shadows on incomplete point-based isosurfaces. Such surfaces are obtained in our setting by a view-dependent isosurface extraction method that allows interactive exploration of large scale datasets on commodity hardware. This approach raises the need for a novel and efficient technique to generate high-quality shadows. The latter dramatically enhances the realism of the visualized isosurface and provides essential cues to perceive depth and small features of the geometry.

Phase: Progressive hardware Assisted Isosurface Extraction:

In this paper[pdf], we introduce a progressive view-dependent isosurface extraction method that exhibits a rapid convergence rate to the exact isosurface and is well suited for remote visualization. The proposed method takes advantage of rendering hardware to resolve visibility tests. In contrast to previous view-dependent isosurface extraction methods, our method Phase can quickly augment the current partial extracted isosurface based on a new point of view without the need for a full view-dependent extraction pass.

View Dependent Isosurface Extraction Methods:

Wise:[pdf] We propose a new approach to polygonal isosurface extraction that is based on extracting only the visible portion of the isosurface. The visibility tests are done in two phases. First, coarse visibility tests are performed in software to determine the visible cells. These tests are based on hierarchical tiles and shear-warp factorization. The second phase resolves the visible portions of the extracted triangles and is accomplished by the graphics hardware.

While the latest isosurface extraction methods have effectively eliminated the search phase bottleneck, the cost of constructing and rendering the isosurface remains high. Many of today's large datasets contain very large and complex isosurfaces that can easily overwhelm even state-of-the-art graphics hardware. The proposed approach is output sensitive and is thus well suited for remote visualization applications where the extraction and rendering phases are done on a separate machines.

Sage:[pdf] We present a dynamic view-dependent isosurface extraction method. Our approach is based on the 3-stage approach we first suggested in WISE, i.e., front-to-back traversal, software pruning based on visibility and final rendering using the graphics hardware. In this work we re-examine the WISE method and similar view-dependent extraction methods, and propose modifications that accelerate the extraction time by a factor of 5 to 10. In particular we suggest using a bottom-up approach for rendering occluding triangles, replacing the traditional marching cubes triangulation with triangle fans and using point-based rendering for sub-pixel sections of the isosurface.

Noise: Near Optimal IsoSurface Extraction:

Noise [pdf] is an algorithm for rapidly extracting isosurfaces from structured and unstructured grids. Using the Span Space, a new representation of the underlying domain, we develop an isosurface extraction algorithm with a worst case complexity of O(sqrt(n)+k) for the search phase, where n is the size of the data set and k is the number of cells in the isosurface. The memory requirement is kept at O(n) while the preprocessing step is O(n log(n)).

The Span Space:

Mantle convection visualization using NOISE:

These are images from two mpeg movies(1.8MB and 4MB) on Mantle Convection Modeling which was done on a Cray T3D supercomputer at the ACL using over 20 million finite elements. The first image shows a single, red, isosurface while the second has two isosurfaces, red and blue, and uses transparency. The third image shows a slice through the mantle with a single isosurface.


  • Problem Solving Environment for Process Data Assimilation

    An Enabling Technology Application Team Collaboration (Part of the PACI initiative).
    Two common challenges that arise in many application areas are:
    (1) how to assimilate massive amounts of measurement data from sensors arrays
    (2) how to process and then use this information to make better decisions.
    The problem arises in such diverse areas as: real-time control of chemical plants, seismic processing, inverse problems in medical imaging, and combinatorial chemistry. A typical example is state estimation where models of the underlying physical and chemical processes are used, together with measurement information, to estimate the state of the system. As more detail is added to the models, and high data rate sensors are used to collect data, there is a need for easy access to Grid-like computational power to make sure that that solutions are available in a time scale that is short relative to the process dynamics. We propose to create a Problem Solving Environment (PSE) for data assimilation that uses new algorithms for Bayesian statistical analysis, combined with sensor data to determine the operating condition of a complex process. We will test the system using process data streams(Mbps) derived from a real chemical plant.
  • Parts, 2D graphics

    Parts where introduced into SCIRun as a new mechanism of building SCIRun modules and connecting them to the user graphical interface (GUI). This mechanism allows automatic assembling of the GUI during runtime.

  • XML Based Integration of Insight Toolkits

    Visualization 2003 tutorial [ppt]
  • Dynamic Templated Code Compilation