Virim: A massively parallel processor for real-time volume visualization in medicine

Günther, Thomas; Poliwoda, Christoph; Reinhart, Christof; Hesser, Jürgen; Männer, Reinhard; Meinzer, Hans-Peter; Baur, Hans Jürgen

doi:10.1016/0097-8493(95)00049-6

Cited by 24 publications

(20 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…This involves a rotation transformation. Straightforward hardware implementation of volume rotation is very expensive [5,7]. Rotation requires global communication and could cause memory contention while writing data back to the distributed memory modules.…”

Section: Shearing On the Cube Architecturementioning

confidence: 99%

Two-Pass Image and Volume Rotation

Chen

Kaufman

2001

Eurographics

View full text Add to dashboard Cite

We present a novel two-pass approach for both 2D image and 3D volume rotation. Each pass is a pseudo shear. However, it has the similar regularity to a pure shear in that a beam remains rigid. Furthermore, the 3D pseudo shear guarantees that beams within one major axis slice remain in the same directional plane after the shearing. These properties make it feasible to implement the pseudo shears on a multi-pipelined hardware or a massively parallel machine. Compared with the existing decompositions, ours offer a minimum number of shears to realize an arbitrary 3D rotation. Our decomposition also preserves the image/volume quality by guaranteeing no minification for the first pass shear.

show abstract

Section: Shearing On the Cube Architecturementioning

confidence: 99%

Two-Pass Image and Volume Rotation

Chen

Kaufman

2001

Eurographics

View full text Add to dashboard Cite

show abstract

“…Second, most work in InfoVis involves data discovery (i.e. finding insight in large datasets and databases) and not data presentation methods: little literature is available on real-time decision making [21][22][23][24] and related work in Defense and Intelligence is mostly classified. Third, InfoVis work has tended to be simplistic, visually naïve and/or highly academic and fail to address real InfoVis demands of real users confronting real time and complex data representation problems.…”

Section: State Of the Art In Scivis And Infovismentioning

confidence: 99%

Architectural Research in Information Visualization: 10 Years after

Bermúdez

Agutter

Foresti

2006

International Journal of Architectural Computing

View full text Add to dashboard Cite

As our civilization dives deeper into the information age, making sense of ever more complex and larger amounts of data becomes critical.This article reports on interdisciplinary work in Information Visualization addressing this challenge and using architectural expertise as its main engine.The goal of this research is to significantly improve real time decision making in complex data spaces while devising a new architecture that responds to complex information environments. Although we have been reporting in aspects of this work for the past 7 years, this paper covers unpublished knowledge, design methods, operational strategies, and other details that bring together all the material published by our group thus far into a comprehensive and useful whole.We conclude by presenting our latest InfoVis design work in Network Security.

show abstract

“…These are disadvantages for hardware implementations. VIRIM [2] is one of typical real-time parallel rendering system in this scheme. It is a parallel rendering system for both parallel and perspective projections, but not organized in a complete parallel-pipeline structure for VLSI implementation.…”

Section: Sample-order Schemementioning

confidence: 99%

“…systolic array, of the method: (1) how to bound a number of voxles to be convoluted and (2) how to organize a convolution. A n umber of voxels to be convoluted in one dimension, M, is computed by: M = 1 + k=k 0 ; (2) where k yy is the slice number or the distance from the base plane and k 0 the distance between the eye position (the center of projection) and the base plane as y A plane the most perpendicular to the viewing vector, which includes the front face of the volume. yy We use notation (x; y; z) for the dataset description and (i; j; k) for that of transformed as shown in Fig.…”

Section: Issues Of Perspective Projectionsmentioning

confidence: 99%

A Parallel Pipeline Convolution for Perspective Projection in Real-Time Volume Rendering.

Ogata¹,

Ohkami²,

Pfister³

et al. 2000

The Journal of the Institute of Image Information and Televisio

View full text Add to dashboard Cite

SUMMARY This paper proposes a convolution with systolic array sturucture for perspective projection in real-time volume graphics based on the shear-warp method. In the original method, the further the ray proceeds, the more voxels are required for the calculation of convolution. The increase of required voxels makes dicult to implement the method in VLSI oriented architecture. 1) We use several sets of resolution of voxels associated with depth, in order that convolution can be done with constant number of voxels independent of depth. 2) We implement 3D convolution by three serial 1D convolutions along X, Y and Z axes, which reduces the calculation steps from M 3 to 3M, where the convolution is calculated for M 3 area. For V 3 voxels dataset, the number of pipelines for rays is V 2 and their pipeline stage is 3M. If the hardware of a single pipeline has the capability of calculating V rays, each of the implemented pipelines is assigned to V theoretical pipelines (for V 2 rays). In actual implementation, a number of hardware pipelines should be much smaller than the V theoretical pipelines. We fold the theoretical pipelines and reduce them to the certain number of hardware pipelines. Regarding this folding, we show the relation between folding process and its necessary time delay. The architecture can generate image of 256 3 voxles dataset( V = 256 ) at 30Hz with 4 pipelines. In addition, the architecture can be extended easily for 512 3 (V = 512) and 1024 3 (V = 1024) dataset with 32 pipelines and 256 pipelines respectively.

show abstract

Virim: A massively parallel processor for real-time volume visualization in medicine

Cited by 24 publications

References 8 publications

Two-Pass Image and Volume Rotation

Two-Pass Image and Volume Rotation

Architectural Research in Information Visualization: 10 Years after

A Parallel Pipeline Convolution for Perspective Projection in Real-Time Volume Rendering.

Contact Info

Product

Resources

About