Region-Restricted Model for Fast Volume Rendering to
Interactively Change Visualized Objects
Michio Oikawa, Koichi Sano, and Yoshiaki Isobe
Systems Development Lab, Hitachi, Ltd., Kawasaki, Japan 215-0013
SUMMARY
With the objective of advanced medical diagnostic
support based on understanding the three-dimensional
structure of the region being diagnosed, volume data meas-
ured by X-ray computed tomography (CT) and three-di-
mensional images obtained by volume rendering are
increasingly being used. To present useful information
about the region being diagnosed, functions to display the
fine structure and to interactively select the information to
be visualized are needed. Interactive operation has been
difficult because the processing time was too long in vol-
ume rendering based on the conventional high-image-qual-
ity image-order approach. Conventional acceleration
techniques to improve interactivity could not be run on thin
clients because of their huge memory consumption and had
problems such as degraded image quality. In this paper, we
propose a region-restricted model which uses the fact that
transparent voxels do not affect the projection plane to
implement fast interactive rendering with low memory
consumption and no image degradation. In addition, we
propose data structures to reduce the accesses to volume
data. The proposed algorithm was applied to real data and
compared to the shearwarp algorithm, which is a conven-
tional acceleration technique. The ability to achieve the
same processing speed while requiring less than one-third
the memory in the calculations was verified. © 2000 Scripta
Technica, Syst Comp Jpn, 31(14): 91100, 2000
Key words: Volume rendering; acceleration; inter-
active process; region-restricted model.
1. Introduction
Recently, in the medical field, diagnostic and thera-
peutic support uses the volume data unit called the voxel
that is measured in three dimensions by X-ray computed
tomography (X-ray CT) or magnetic resonance imaging
(MRI) [13].
To provide diagnostic support by presenting realistic
three-dimensional structure information about the diagno-
sis region from a large volume of measurement data, the
measurement data must be visualized in three dimensions.
Since not only surface information about the measurement
subject but a variety of information about internal organs
are contained in the measurement data, the ability to inter-
actively select useful information for diagnosis and to dis-
play high-image-quality images visualizing the fine
structures is important.
Volume rendering is suited to implementing these
kinds of functions. High-image-quality three-dimensional
images can be generated from ambiguous measurement
data obtained at object boundaries in X-ray CT. Addition-
ally, information about the organ or tissue being displayed
can be changed by changing the opacity settings of the
voxels.
Previously proposed volume rendering algorithms
can be classified based on the algorithm into the following
categories: image-order approach, object-order approach,
and hybrid approach [4]. The object-order approach and the
hybrid approach are techniques for speeding up the image-
order approach. Each one features either the image quality,
the processing speed, or the balance between the image
quality and the processing speed and has advantages and
© 2000 Scripta Technica
Systems and Computers in Japan, Vol. 31, No. 14, 2000
Translated from Denshi Joho Tsushin Gakkai Ronbunshi, Vol. J82-D-II, No. 1, January 1999, pp. 127136
91