 |
Description  |
|
|
BACKGROUND OF THE INVENTION
In a wide variety of modern applications, it is desirable to observe the
three dimensional coherence of a volume or object of interest. In the case
of imaging real three dimensional ("3D") solids, ordinarily it is only
possible to view discrete planar cross sections of the 3D volume and its
contents. It is not possible typically to view 3D volumes such that
internal and object surfaces, boundaries, interfaces, and spatial
relationships within the volume can be separated and identified visually.
In medical imaging, for example, it is highly desirable to be able to
visualize anatomical structures by three dimensional representations on
computer graphic displays. The ability to produce accurate, volumetric,
anatomical models from computerized tomographic (CT) scans is extremely
valuable as a surgical aid, (such as for displaying structures to plan
surgical intervention, or to represent details of the anatomical structure
without the need for exploratory surgery). Thus, the 3D shape, size and
relative position of pathologic structures provides important data for
surgical planning, diagnosis and treatment.
Computer simulation of real 3D volumes depends on the ability to
reconstruct 3D structures from planar section data, such as CT scans.
These and other scans provide data from which a 3D density image volume
consisting of volume elements (voxels) is available as input data for
image processing. Unfortunately, such input image volumes are typically of
low resolution as compared to the level of detail desired to represent
accurately the sampled volume.
For example, in CT scan image volumes, voxels represent x-ray attenuation
or other image volume data throughout the volume, including across surface
boundaries. Although a voxel is assigned only a single "homogenous" value,
in fact there exists a boundary and discrete materials on either side of
the boundary within the object under consideration. Thus, a voxel along an
edge is a sample extending over both sides of the edge. Further, if a
material (such as a bone) is less that one voxel wide, the voxel that
provides boundary information about that bone is very low resolution.
Thus, the boundary shape within a voxel is not readily determined.
Various approaches have been used in an effort to approximate surface
boundaries within volumes. One well known method is "thresholding". In
thresholding, voxels that cross a boundary are classified as being
composed of one or the other material type on either side of the boundary.
The projected visible boundary thus becomes the binary classified voxel
border.
The larger the voxels, the greater the error that is introduced by
thresholding. Further, for coarse images or images with high density and
closely spaced surface boundaries, thresholding provides an even further
degraded result, such that the resultant images become less and less
accurate. Subsequent approximation techniques are sometimes used in an
attempt to render a more accurate approximation from the thresholding
result. However, attempts to approximate the unknown surface gives rise to
a grossly inaccurate result since these approximations rely on the initial
binary classification of the voxels.
It is also highly desirable to be able to view all the data within the
volume simultaneously from selected stationary or rotating views; that is,
to view into the center of a volume, and to detect objects and hidden
surfaces within the volume (and therefore internal boundaries and
surfaces). To do so, it is necessary to be able to see partially through
interfering objects when desired (e.g., for bone surrounded by muscle, to
be able to observe the bone as well as the muscle surrounding the bone).
Prior art techniques for rendering volume elements force a binary decision
to be made as to whether a pixel is made of a given material or not. A
binary classification introduces aliasing (or sampling) errors as the
continuous image function is not preserved. The error introduced by binary
classification is introduced upon any attempt to reconstruct the original
image volume from the classified volume. Because the reconstructed image
can only have as many intensity levels as there are materials, material
interfaces will be jagged and the intensities will not represent the
original image function.
Further, it may be desirable to shade the volume data so that boundaries
and surfaces may be determined more accurately. In addition, it may be
desired to shade the surface of a projected image to give the impression
that the image is illuminated by one or more sources of light. Such
surface shading can also be used to emphasize and highlight selected areas
of the displayed image volume.
SUMMARY OF THE INVENTION
The imaging system of the present invention provides apparatus and methods
for projecting a two dimensional (2D) representation of three dimensional
(3D) volumes where surface boundaries and objects internal to the volumes
are readily shown, and hidden surfaces and the surface boundaries
themselves are accurately rendered. Also, the two dimensional image
produced is capable of having the same resolution as the sampling
resolution of the input image volume of interest. This is accomplished
through the implementation methods for determining "partial volumes" of
voxels. Partial voluming provides for the assignment of selected colors
and opacities to different materials (or data components) represented in
an image data volume based on the percentage composition of materials
represented in each voxel of the image volume. Unlike prior art systems,
such as those that use thresholding techniques, the imaging of the present
invention has a high degree of precision and definition that is
independent of image volume on a per voxel basis.
An image volume representing a volume object or data structure is written
into picture memory. A color and opacity is assigned to each voxel within
the volume and stored as a red (R), green (G), blue (B), and opacity (A)
component, three dimensional data volume. The RGBA assignment for each
voxel is determined based on the percentage component composition of the
materials represented in the volume, and thus, the percentage of color and
transparency (based on 100% reference material color and transparency
values) associated with those materials. Such voxels stored within the
picture memory for the component channel volume are sometimes referred to
herein as RGBA voxels.
Next, the voxels in the RGBA volume are used as mathematical "gels" or
filters such that each successive voxel filter is overlayed over a prior
background voxel filter. Through a linear interpolation, a new background
filter is determined and generated. The interpolation is successively
performed for all voxels (or groups of voxels, e.g., voxel scan lines) up
to the front most voxel for the plane of view. The method is repeated
until all display voxels are determined for the plane of view.
The present inventive method of reconstruction provides for the 2D
projection of volumes where surfaces of objects within a discrete data
volume can be detailed and displayed even where surfaces could not
previously be found because image volume data is course or the sampling
rate is low. The present invention provides for visualization of hidden
surfaces within the volume such that all surface boundaries are clearly
and accurately defined.
Thus, given a discrete sample data volume of a complex volume data set or
object, and finding the percentage of composition of component materials
for each voxel, it is possible to render complex images without initially
being provided with specific information as to boundary location within
each voxel. The present inventive method and apparatus thereby provides
for viewing volumes without having a geometric model (but rather having
only a 3D raster volume) such that it is possible to create a complex
projection of the 3D volume, using partial transparency and color that can
have the characteristics of a mathematical geometric solid model. Objects
can be viewed which partially obscure other objects, and spatial
relationships between objects can be accurately rendered.
The present inventive method of shading provides for the rendering of
surfaces and boundaries to subvoxel accuracy. A three dimensional gradient
at the center of each RGBA voxel is computed based on the opacity of
neighboring voxels. The gradient volume consists of XYZL voxels, where the
X,Y and Z components represent the change in opacity in each direction,
and L is the length of the amplitude of the gradient. By multiplying each
RGBA voxel by the length of its gradient, surfaces remain, but solid
regions become more transparent.
In order to shade the surface of the displayed image volume, a source of
light is chosen as being at a fixed point in space relative to the image
volume. Light vectors L are defined extending from the light source to the
center of each voxel in the image volume. A shading function is generated
based on the angle of the gradient vector of each voxel and the light
vector associated with each voxel. In the preferred embodiment of the
present invention, the shading function includes user determined variables
which allow for the addition of back lighting and side lighting to provide
additional surface highlighting.
Other objects and advantages of the present invention will become more
apparent upon a reading of the Detailed Description of the Invention in
connection with the drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a block diagram illustrating the architecture of the present
invention.
FIG. 2 is a block diagram illustrating the division of the picture memory
of the present invention.
FIG. 3 is block diagram illustrating the architecture of the channel
processor of the present invention.
FIG. 4a is a graph showing a typical input data histogram.
FIG. 4b is a graph illustrating classification of a CT input data
histogram.
FIG. 5 illustrates a plurality of voxels used to generate a two dimensional
display.
FIG. 6 illustrates image reconstruction for successive voxel scan lines and
works.
FIG. 7 illustrates the calculation of projected pixel output based on the
concatenated filtering of the present invention.
FIG. 8 illustrates the method of generating a gradient vector for a
selective voxel in two-dimensional space.
FIG. 9 illustrates a shaded image section, shaded in accordance with the
present invention.
FIG. 10 illustrates the generation of a gradient vector for a selected
voxel in three dimensional space.
FIG. 11 illustrates the surface shading effect of the present invention.
FIG. 12 is a flow chart setting forth the steps in implementing the present
invention.
DETAILED DESCRIPTION OF THE INVENTION
Notation and Nomenclature
The detailed description which follows is presented largely in terms of
algorithms and symbolic representations of operations on data bits within
a computer memory. The algorithmic descriptions and representations are
the means used by those skilled in the data processing arts to most
effectively convey the substance of their work to others skilled in the
art.
An algorithm is here, and generally, conceived to be a self-consistent
sequence of steps leading to a desired result. These steps are those
requiring physical manipulations of physical quantities. Usually, though
not necessarily, these quantities take the form of electrical or magnetic
signals capable of being stored, transferred, combined, compared and
otherwise manipulated. It proves convenient at times, principally for
reasons of common usage, to refer to these signals as bits, values,
elements, symbols, characters, terms, numbers, or the like. It should be
kept in mind, however, that all of these and similar terms are to be
associated with the appropriate physical quantities and are merely
convenient labels applied to these quantities.
Further, the manipulations performed are often referred to in terms, (such
as adding or comparing) which are commonly associated with mental
operations performed by a human operator. No such capability of a human
operator is necessary, or desirable in most cases, in any of the
operations described herein which form art of the present invention; the
operations are machine operations. Useful machines for performing the
operations of the present invention include general purpose digital
computers or other similar devices. In all cases the distinction between
the method of operations and operating a computer, and the method of
computation itself should be noted. The present invention relates to
methods for operating a computer in processing electrical or other (e.g.,
mechanical, chemical) physical signals to generate other desired physical
signals.
The present invention also relates to apparatus for performing these
operations. This apparatus may be specially constructed for the required
purposes or it may comprise a general purpose computer as selectively
activated or reconfigured by a computer program stored in the computer.
The algorithms presented herein are not inherently related to any
particular computer or other apparatus. In particular, various general
purpose machines may be used with the teachings herein, or it may prove
more convenient to construct more specialized apparatus to perform the
required method steps.
While voxels are for purposes of convenience sometimes referred to as three
dimensional elements having a three dimensional volume, it should be
appreciated that voxels define points in a three dimensional space.
In addition, the following description, numerous details are set forth such
as algorithmic conventions, specific numbers of bits, etc., in order to
provide a thorough understanding of the present invention. However it will
be apparent to one skilled in the art that the present invention may be
practiced without these specific details. In other instances, well-known
circuits and structures are not described in detail in order not to
obscure the present invention unnecessarily.
System Architecture
The system architecture of the present invention is illustrated in FIG. 1.
In the preferred embodiment, the image system comprises a host computer 10
coupled to a bit map terminal 12 and a system bus 14. The system bus 14 is
coupled to a channel processor 16, a memory controller 18, and a video
controller 20. The memory controller 18 is coupled to the channel
processor 16 and video controller 20 and to a picture memory 26. The video
controller 20 is coupled to the memory 26 via a video bus 28. Channel
processor 16 is coupled to the picture memory 26 via a picture bus 24. A
color monitor 22 or other ouput device is coupled to video controller 20.
The channel processor 16 comprises a parallel processing computer system.
In the preferred embodiment of the present invention, four sixteen bit
parallel processors are utilized in the channel processor. Although it has
been found advantageous to utilize four parallel processors in the present
invention, it will be obvious that any number of processors may be used
without departing from the spirit and scope of the present invention.
The host computer 10 is coupled to a bit map terminal 12 which displays a
plurality of menus available to the user of the present invention. In the
preferred embodiment, the host computer 10 is utilized to manipulate the
histogram generated from the image data stored in the picture memory 26. A
look up table may be stored in the host computer 10 containing preassigned
color and opacity values for each peak intensity location in a histogram.
In addition, the host computer 10 may be utilized to display the histogram
on bit map terminal 12 so that user defined color and opacity values may
be determined.
In the present invention, the system bus 14 is a sixteen bit bus coupling
the host computer 10 to the other components of the image processing
system. Although sixteen bit parallel processors are utilized in the
present invention, eight bit, thirty two bit, or any other size processor
may be utilized as well. Thus, other bit size system buses may be utilized
as required.
The video controller 20 provides vertical and horizontal synchronization to
the color monitor 22 and also includes a buffer memory for displaying
screen data. The video controller 20 accesses the video bus 28 during
vertical blanking periods to provide updated information to the buffer
memory. The video controller includes a digital to analog (d/a) converter
to convert digital signals into a video signal suitable for display on the
monitor 22. In addition, the controller 20 includes look up tables for
each of the RGB channels. The twelve bit digital word of the red (R)
channel, for example, is mapped out through a look up table and a unique
red color value is outputted and displayed upon color monitor 22. As
noted, there are look up tables for each of the channels R, G and B. The
video controller 20 functions as a three value in--three value out
converter, which allows gamma corrections. The video bus 28 in the
preferred embodiment is a bus divisible in 12 bit units and operating at
up to least 420 megabytes per second. The picture bus 24 in the preferred
embodiment operates at 240 megabytes per second.
The color monitor 22 of the present invention may be any commercial color
monitor having 480 to 1125 number of scan lines. The present invention has
the speed and capability to service high resolution, high scan line
monitors.
The picture memory 26 is a random access memory (RAM) utilized for storing
the original image data as well as classification data, color and opacity
data and the composite RGBA volume data. The picture memory 26 is shown in
detail in FIG. 2. As previously noted, each voxel is represented by 4
twelve bit digital words. One portion of the memory, original volume
memory 26a, stores the data in its original form. In the present preferred
embodiment, approximately 40 million bits are dedicated for storage of
original data. Original image volume data is stored as a series of twelve
bit words where each word represents the intensity level of one voxel of
the original voxel volume. After the RGBA volumes are generated, they are
stored in RGBA volume memory 26b. Each color value is represented by a
twelve bit word and the opacity value A is also represented by a twelve
bit word. Thus, each voxel in the RGBA volume memory is represented by 4
twelve bit words. In the preferred embodiment of the the present
invention, 150 millions bits of memory are dedicated to the RGBA volume
memory 26b. The composite voxels are stored in the two-dimensional
projection memory 26d. As noted previously, the combination voxels are
generated from concatenation of various scan lines of the RGBA volume
data. The picture memory 26 also includes temporary work space memory 26c,
RGBA shaded volume memory 26e and XYZL gradient volume memory 26f.
Although the number of bits in various portions of the picture memory 26
have been described, it will be obvious to one skilled in the art that any
suitable number of bits of memory may be utilized in practicing the
present invention.
Referring again to FIG. 1, memory controller 18 is utilized to arbitrate
all accesses to the picture memory 26. Memory controller 18 enables data
to be written into the picture memory 26 through means of picture bus 24.
Channel processor 16 is utilized to generate the histogram from which
classifications of the voxels is made. Channel processor 16 is shown in
detail in FIG. 3, and comprises scratch pad memory 17 and four
multiplier/arithmetic logic units (ALU) 15a through 15d, respectively. In
the preferred embodiment, scratch pad memory 17 includes 64K sixteen bit
word memory locations. Channels processor 16 utilizes the scratch pad
memory 17 for temporary storage of data for manipulation by the
multiplier/ALU pairs. Each of the multiplier/ALU pairs 15a through 15d is
dedicated to one of the four co-processors that comprise the channel
processor. This parallel approach allows the same instruction to be
executed on different data. Such a system is referred in the art as a
single instruction, multiple data stream (SIMD) system. One such inventive
system is more fully described in U.S. patent application Ser. No.
748,409, filed June 24, 1985, and owned by the assignee of the present
invention. In the preferred embodiment, one channel is utilized for the R
(Red) values of the voxels, one channel is utilized for the G (Green)
values, one for B (Blue) and one for A (opacity) values.
The initialization of the processing system occurs when the original image
volume data is read into the host computer 10. The original image data may
be transmitted in a compressed form and require decoding and expansion. In
addition, depending on the source of the data, the bits/pixel resolution
of the raw data may not match the preferred twelve bit/pixel resolution of
the present invention. Therefore, if required, the host computer or
channel processor may artificially enhance the original image data to an
acceptable resolution.
The host computer 10 then outputs a request signal on the system bus 14
requesting the channel processor 18 to allow the original image data to be
written into the picture memory. The host computer 10 then outputs the
original image data onto the system bus 14 into the channel processor 16.
The channel processor 16 then outputs the original image data onto the
picture bus 24 where it is written into the original image volume memory
26a of the picture memory 26.
If desired, or necessary the original image volume is then classified by
the channel processor 16. A scan line from the original volume memory 26a
is then placed on the picture bus 24 and inputted to the channel processor
16. The channel processor 16 counts the number of voxels at each intensity
level of the 2,049 definable intensity levels to generate a histogram, the
channel processor 16 may then classify the peaks pursuant to previously
defined look up tables or a program hierarchy. Additionally, the histogram
may be displayed by the host computer 10 on the bit map terminal 12 so
that the classification may be manually performed by a user.
After classification, the classification data is sent through system bus
14, to the channel processor 16, and in particular to the scratch pad
memory 17. The channel processor computes a plurality of lookup tables
based on classification data from the host. The scratch pad memory 17
contains a plurality of look up tables so that RGB and opacity values may
be assigned to the classified voxels. The channel processor 16 requests
one scan line (a one dimensional array of voxels) at a time from the
memory controller and processes the scan line through the stored look up
tables. The input to the channel processor look up tables is a monochrome
scan line and the output is an RGBA classified scan line. The scratch pad
memory 17 of the channel processor 16 includes at least three buffers,
namely a scan line buffer, a look up table, and an output buffer for
output classification. Although described in conjunction with look up
tables, any suitable method of assigning or generating color values may be
utilized in the present invention.
The output from the output buffer is coupled through a picture bus 24 to
the RGBA volume memory 26b of picture memory 26. Each slice of the
original image volume is stored as an array in the RGBA volume memory 26b.
The "concatenated filtering" of the present invention is performed by the
channel processor 16. A background image volume is initially read into the
scratch pad memory of the channel processor one scan line at a time. As
each succeeding image volume is read into the channel processor 16, the
RGBA values are concatenated by the multiplier/ALU Units 15a-d
respectively. This composite image volume is stored in the projection
memory 26d of picture memory 26.
To display the composite image, the contents of memory 26d is outputted
onto the video bus 28 into the video controller 20. The digital data
stored in the projection memory 26d is converted to video analog data by
the controller 20 and outputted to color monitor 22 where it is displayed
in a raster scanning fashion.
Partial Volumes
Classification
An image volume, that is, a volume of picture elements (voxels)
representing a three dimensional image, is read into host computer 10 and,
after any required decoding or decompression, into original image memory
26b of picture memory 26. Thus, the object(s) under consideration, an
array of two dimensional arrays, can be thought of and mathematically
represented as as a three dimensional ordered array in picture memory. The
image volume may be an unprocessed image array that may be obtained by
various methods known in the art.
While the present invention has wide application, it is described for
purposes of example in connection with the three dimensional display of
computed tomographic ("CT") images. Input image volume data, represented
as an ordered CT image volume array in this context, may be obtained from
a variety of tomographic imaging methods, e.g., x-ray computed tomography,
ultrasound sector scanning, nuclear magnetic resonance, etc. In other
contexts, the image volume input data may be obtained using other imaging
methods, such as seismic imaging, or the result of computer model or
simulation, (such as a fluid flow simulator), for example. The image
volume in the present example is stored as an ordered array of 2D images
where each image is a 2D ordered array of 12 bit numbers.
The CT scan or other image volume provides monochromatic grey scale input
data to the image processing system of the present invention. This CT
input image volume data is stored in the initial volume memory 26(a) of
picture memory 26 as an ordered array of 12 bit binary numbers, each
representing a given CT scan or other image volume data (voxel). In the
present example, the image volume information which is provided by the CT
scan input represents information about four discrete materials in the
anatomy under consideration; namely, air, fat, soft tissue, and bone.
In the present example, the intensity of the grey scale data depends on the
x-ray density of the material represented from the original imaging
method. Referring to FIG. 4(a), grey scale intensity data for each voxel
is plotted in a histogram 30 that provides a distribution of the number of
voxels in the image volume versus intensity. The histogram 30 is generated
by the channel processor 16 from image volume input data and is
transmitted to the host computer 10 via system bus 14.
The grey scale level histogram 30 is a function which shows, for each grey
level, the number of voxels in the image volume that have a particular
grey level. The abscissa 32 is grey level intensity, shown in this case to
be from 0 to 2048. The ordinate 34 is frequency of occurrences (e.g.,
number of voxels of the image volume at each intensity). Thus, the area
under the histogram curve 36 represents the total number of voxels in the
image volume.
The histogram 30 is derived from a particular image which, in the present
example, corresponds to the twelve-bit pixel information of the CT scan.
The resulting histogram is then filtered by the host computer 10 to remove
noise, (i.e., to supress high-frequency variations while preserving the
shape of the input function), using low pass filtering techniques widely
known in the art and results in a smoothed histogram curve as shown in
FIG. 4a.
In the prior art imaging systems, all voxels are classified as representing
100% of single material. Thus, using prior art techniques, a binary
classification is made for each voxel, in which all voxels within a
particular intensity region are classified as one or the other of those
materials represented by the input volume image data. Using prior art
techniques for the present CT scan example, voxels would be classified as
either bone, fat, tissue or air. In the prior art, a threshold grey level
is chosen as a cutoff at some point between histogram peaks. All voxels
within a given intensity range are thereby categorized either 100% air,
soft tissue, bone or fat. This information is then stored in memory as a
two bit or one bit value.
For CT renderings, this binary classification is suitable for peak regions
of the histogram where a voxel content clearly falls within one discrete
material classification or another. However, such "hard" classification
requires that a discrete binary determination be made concerning the
material classification of all voxels within the image volume.
For example, in the prior art, at surface boundaries (such as at an
interface where tissue is attached to bone), a voxel crossing an edge
location is classified as either bone or tissue. Thus, where voxels cross
between surfaces or along edges, or where the local material is less than
1 voxel wide, the edge or surface boundary will be lost, and an imprecise
rendering will result. Such binary value assignment, therefore, introduces
significant aliasing error, resulting in a surface rendering that is
inaccurate.
In the present invention, voxels are classified, in accordance with their
associated intensities, as being composed of between 0% and 100% of each
material type represented in the image volume.
Referring to FIG. 4b, significant peaks P and troughs T are identified in
the histogram 30. An intensity range 40 characterized as 100% of a
material is determined for all voxels falling within a selected intensity
range to the left and right of significant histogram peaks P. In the
present embodiment for a CT scan, this range is determined by a linear
approximation. The 100% material range is approximated as all voxels P
falling within the intensity band defined as .vertline.p-t.vertline./4
where p is the intensity associated with a significant peak P and t is an
intensity associated with an adjacent significant trough T. All voxels
within this range on either side of a given significant peak P are
classified as being a 100% "pure" material value. Thus, in the present
example, all voxels within these ranges are categorized as 100% fat, air,
soft tissue, and bone, respectively. Voxels outside these intensity range
values are categorized as containing image information representing some
percentage of the "pure" materials to the left and to the right of the
voxels. Therefore, they are combination material voxels each having a
partial percentage composition of these materials.
For convention, voxels classified as being composed of less than 100% of
any single material are referred to as partial volumes in connection with
the present invention. Further, voxels classified as 100% of one material
are sometimes referred to as such. However, in more general terms,
utilizing the present invention, classification is based on partial volume
determination. Thus, as a more general convention, all classified voxels
are "partial volumes", where voxels classified as a single homogenous
material are "partial volumes" having 100% of one material, and 0%
composition of other component materials.
In determining partial volume classification, the percentage of each
material contained in voxels falling outside the 100% material intensity
ranges is found. In the present preferred embodiment, the percentage voxel
material content is determined by performing a linear interpolation for
voxels having intensities falling between the limits of adjacent 100%
material intensity values. A ratio is then found of: (1) the intensity
differences from a 100% value boundary to the voxel intensity under
consideration (numerator) versus (2) the intensity range value between
surrounding 100% material value limits (denominator). This ratio provides
the percentage voxel composition for each adjacent material at a
particular intensity value between 100% value ranges.
Referring now to FIG. 4b, partial volume percentages are determined by the
channel processor calculation performed for all partial volume voxels by a
linear interpolation for each intensity between adjacent 100% material
intensity value limits. Alternatively, classification may be made pursuant
to values located in a look-up table or made pursuant to inspection and
user determined values. In the present preferred embodiment, for a given
voxel associated with partial volume intensity I(pv), its percentage
composition is:
##EQU1##
where PVab is the partial volume voxel(s) under consideration;
I(PV)ab=intensity of partial volume voxels under consideration between
100% material intensity regions a and b; I(Pb)a=intensity at peak boundary
value a for 100% material type; and I(Pb)b=intensity at boundary value b
for 100% material type;
When calculating the fraction of material associated with boundary a,
I(Pb)a is used; with boundary b, I(Pb)b is used. Thus, each voxel is
classified as either; (1) a partial volume made up of a percentage of more
than one image volume material type, or (2) a "pure" volume made up of
100% of one image volume material type.
In the present preferred embodiment, the classification of grey scale data
is performed for a musculo-skeletal CT scan model. In this model, as shown
in FIG. 4b, the relationship of material grey scale intensities is such
that fat 40 is always adjacent to soft tissue 44; soft tissue 44 always
surrounds fat 40 and also always surrounds bone 46. It should be
appreciated, however, that other classification schemes can be used to
determine percentage voxel material composition for other input data, such
as for example, seismic image volume input data. Also, such classification
may be used for applications where input data includes material
adjacencies in which multiple materials (and/or their associated histogram
intensities) may be concurrently adjacent to each other and are so
represented by image volume input data. Further, although straight line
interpolat | | |
