Survey of Parallel Volume Rendering Algorithms

1
Survey of Parallel Volume Rendering Algorithms
by Craig M. Wittenbrink craig_wittenbrink_at_hpl.hp.c
om Visual Computing Department Hewlett-Packard
Laboratories
presentation by Nicholas Schwarz schwarz_at_evl.uic.e
du Electronic Visualization Lab University of
Illinois at Chicago
2
Introduction

• Visualizing large volume data is computationally
  complex O(n3) for an n x n x n volume.
• Use parallelism to achieve interactivity.
• Current classifications make it difficult to
  determine the best approach.
• Presents a 5 class taxonomy for parallel volume
  rendering algorithms.

3
Volume Rendering Taxonomy

• Algorithm Control Flow
• View Reconstruction
• Backward Ray Casting
• Forward Splatting
• Multipass Forward
• Fourier
• Outer Loop Data Space
• Object Space
• Image Space

4
Volume Rendering Taxonomy

• Targeted Hardware
• Graphics (G)
• Volume Rendering (VR)
• Parallel Shared Address Space (PS)
• Parallel Distributed Address Space (PD)
• Distributed (D)

5
Volume Rendering Taxonomy

• Application Data Characteristics
• Input Topologies
• Rectilinear (R)
• Curvilinear (C)
• Unstructured (U)
• Data Types
• Scalar, Vector, Tensor
• Data Units
• Voxel Format

6
Volume Rendering Taxonomy

• Visualization Method
• Shading, Data Classification, Transmission Model,
  Reconstruction Filter, Gradient
• Publication Specifics
• Date, Number of Processors, Volume Size,
  Prototype Machine, Asymptotic Complexity
  Analysis, and Performance

7
Survey
8
Links

• http//www.hpl.hp.com/techreports/98/HPL-98-49R1.h
  tml