Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
INTRODUCTION:
Higher dimensional spaces have since become one of the foundations for formally
expressing modern mathematics and physics. Large parts of these topics could not exist in their
current forms without the use of such spaces. Einstein’s concept of Space-time uses such a 4D
space, though it has a Minkowski structure that is a bit more complicated than Euclidean 4D
space.
4D visualization takes three-dimensional images and adds the element of time to the
process. The revolutionary capabilities of new three-dimensional (3-D) and four-dimensional (4-
D) medical imaging modalities along with computer reconstruction and rendering of
multidimensional medical and histologic volume image data, obviate the need for physical
dissection or abstract assembly of anatomy and provide powerful new opportunities for medical
diagnosis and treatment, as well as for biological investigations.
When we bring the cube into the fourth dimension, we begin to experience some
counterintuitive math. We extrude the cube in a direction perpendicular to all of the first three.
This is impossible within the third dimension because there are only 3 dimensions which the
cube is already expanded in. When we add the fourth dimension, in order to maintain the
properties of the cube of all angles being 90 degrees and all sides being the same, we must
extrude in this new dimension.
Fig-1.1
Cubes in the fourth dimensions are technically called tesseracts. Objects in 4D differ in
length, width, height, and trength. Superimposing trength on any of the previous dimensions
gives an object in the subsequent dimensions a trength of 0, or a value that is infinitely small.
All of the edges of a tesseract are the same, and all of the angles are right. This makes
sense in theory, but when we begin to imagine what a tesseract would look like, we are bound by
our 3-dimensional minds. To view a tesseract, we have to superimpose this fourth-dimensional
object into the third dimension.
Locked within 3-D biomedical images is significant information about the objects and their
properties from which the images are derived. Efforts to unlock this information to reveal
answers to the mysteries of form and function are couched in the domain of image processing
and visualization. A variety of both standard and sophisticated methods have been developed to
process (modify) images to selectively enhance the visibility and measurability of desired object
features and properties. For example, both realism-preserving and perception-modulating
approaches to image display have significantly advanced the practical usefulness of 4-D
biomedical imaging.
Concept of 4D Visualization
In the field of scientific visualization, the term "four dimensional visualization" usually
refers to the process of rendering a three dimensional field of scalar values. While this paradigm
applies to many different data sets, there are also uses for visualizing data that correspond to
actual four-dimensional structures. Four dimensional structures have typically been visualized
via wire frame methods, but this process alone is usually insufficient for an intuitive
understanding. The visualization of four dimensional objects is possible through wire frame
methods with extended visualization cues, and through ray tracing methods. Both the methods
employ true four-space viewing parameters and geometry.
The ray tracing approach easily solves the hidden surface and shadowing problems of 4D
objects, and yields an image in the form of a three-dimensional field of RGB values, which can
be rendered with a variety of existing methods. The 4D ray tracer also supports true four-
dimensional lighting, reflections and refractions. The display of four-dimensional data is usually
accomplished by assigning three dimensions to location in three-space, and the remaining
dimension to some scalar property at each three-dimensional location. This assignment is quite
apt for a variety of four-dimensional data, such as tissue density in a region of a human body,
pressure values in a volume of air, or temperature distribution throughout a mechanical object
RAY TRACING ALGORITHM:
4D IMAGE WARPING
METHOD:
(1) Rigid alignment of 3D images of a given subject acquired at different time points, in
order to produce a 4D image. 3D-HAMMER is employed to establish the correspondences
between neighboring 3D images, and then align one image (time t) to its previous-time image (t-
1) by a rigid transformation calculated from the established correspondences.
(2) Hierarchical deformation of the 4D atlas to the 4D subject images, via a hierarchical
attribute-based matching method. Initially, the deformation of the atlas is influenced primarily by
voxels with distinctive attribute vectors, thereby minimizing the chances of poor matches and
also reducing computational burden. As the deformation proceeds, voxels with less distinctive
attribute vectors gradually gain influence over the deformation
Some believe that it is impossible for us to visualize 4D, since we are confined to 3D and
therefore cannot directly experience it. However, it is possible to develop a good idea of what
4D objects look like: the key lies in the fact that to see N dimensions, one only needs an (N-1)-
dimensional retina.
Even though we are 3D beings who live in a 3D world, our eyes actually only see in 2D.
Our retina has only a 2D surface area with which it can detect light coming into our eye. What
our eye sees is in fact not 3D, but a 2D projection of the 3D world we are looking at.
In spite of this, we are quite able to grasp the concept of 3D. Our mind is quite facile
at reconstructing a 3D model of the world around us from the 2D images seen by our
retina. It does this by using indirect information in the 2D images such as light and
shade, parallax, and previous experience. Even though our retina doesn't
actually see3D depth, we instinctively infer it. We have a very good intuitive grasp of
what 3D is, to the point that we are normally quite unconscious of the fact we're only
seeing in 2D.
Similarly, a hypothetical 4D being would have a 3D retina, and would see the 4D
world as 3D projections.
It would not directly see the 4th dimension, but would infer it using indirect
information such as light and shade, parallax, and previous experience.
The key here is that what the 4D being sees in its retina is 3-dimensional, not 4-
dimensional. The 4th dimension is inferred. But since we have a good intuitive grasp
of 3D, it is not that difficult to understand what a 4D being sees in its retina. From
there, we just need to learn how to infer 4D depth.
The rest of this document will describe in detail the basic principles of 4D
visualization, as well as provide a number of examples of 4D objects. We shall take a
purely geometrical approach and treat all 4 dimensions as spatial dimensions.
Dimensional Analogy
Let's begin with the very basics. Let's start in a 1D world. The 1D world is like
a piece of string. There is only one axis along which one may move along this string,
the X-axis. The only dimension any object can have is length, because there aren't any
other dimensions to accomodate width or height. So the only possible objects in 1D
are points, which are 0D, and lines, which are 1D.
Now, let's move to the next higher dimension. The 2D world is a plane, like the
surface of a piece of paper, but extending indefinitely in width and length. The 2D
world is much more interesting than the 1D world, because a much larger variety of
objects are possible. For example, we can have polygons and circles, in addition to
points and lines:
Now let's move on to the 3D world. Objects in the 3D world are bounded not by lines
or curves, but by 2D surfaces. For example, a cube is bounded by 6 squares, and a ball
is bounded by a spherical surface. The spherical surface is 2D, because any point on
the sphere is fully specified by only two parameters: longitude and latitude.
We can see a pattern emerge here. Objects in 1D are bounded by 0D points; objects in
2D are bounded by 1D lines (or curves); and objects in 3D are bounded by 2D
surfaces. In other words, points in 1D are analogous to lines and curves in 2D: they
form the boundaries of objects in the respective dimensions. Similarly, bounding lines
and curves in 2D are analogous to surfaces in 3D. So, by applying dimensional
analogy, we see that in N dimensions, objects are bounded by (N-1)-dimensional
boundaries.
This leads us to conclude that in 4D, objects are bounded not by points, lines, nor
even surfaces, but volumes. It would be rather difficult to realize this without applying
dimensional analogy. For example, as we shall see later, a 4D cube is bounded by 8
cubes. We call these bounding volumes the cells of the 4D cube.
Vision
The reason is that in order for us to see something, light must have
an unobstructedpath from that thing to the cells in our retina. A 2D retina works,
because there is a 3rd dimension in which the light can travel unobstructed from the
object onto the retina. However, if our retina were 3D, we would not see anything
more, because light must pass the cells on the outer surface in order to reach the inner
cells, so that what the inner cells see has already been seen by the outer cells. Since
we are confined to 3D, there is no additional dimension in which light may travel to
reach these inner cells by an independent path, which might have given us additional
visual information.
Projections (1)
We can immediately see where the corners of the cube are, and that its faces are
tetragonal. We can see that each corner has three faces meeting at it. Now look again
at the sequence of cross-sections we saw in the previous chapter:
Can you pick out where the corners of the cube are in these cross-sections? What
about the shape of the cube's faces? How many faces meet at each corner? We would
have to analyse these cross sections very carefully in order to derive this information.
However, using projections this information is immediately available. Although the
cross-section method does give us valuable information, the projection method is
much easier to understand.
Types of Projections
There are two main categories of projections: parallel projection and perspective
projection. In parallel projection, the projection rays are parallel to each other. In
perspective projection, the rays converge on a single point behind the projection
plane. Both types of projection are useful in examining higher-dimensional objects.
Parallel Projection
Parallel projection may be further classified into orthographic projection, where the
rays are always perpendicular to the projection plane, and oblique projection, where
the rays intersect the projection plane at an angle.
The image of the cube under orthographic projection is a square, because we are
looking at it face-on.
Just as the image of the cube under orthographic projection is a 2D square, so the
image of the hypercube in orthographic projection is a 3D cube.
We have colored the nearest cell of the hypercube red, and the farthest cell blue, and
the remaining cells yellow. Note the similarity between this image and the oblique
projection of the 3D cube. Just as the oblique projection of the cube looks like two
squares connected by 4 lines, so the oblique projection of the hypercube looks like
two cubes connected by 8 lines. You can see dimensional analogy at work here.
Some of the properties of images produced by parallel projection are:
The size of the image does not depend on the distance of the object. No matter
how far away the object is, the parallel rays will always produce the same
image size.
Parallel lines in the object remain parallel in the image. For example, in the
oblique projection of the cube, the vertical edges of the cube all appear vertical
in the image. The front-to-back edges appear slanted, but they are still parallel
to each other, just as in the real cube.
Projections (2)
Perspective Projection
Parallel projections, while useful, lack an important visual cue: distance. Since
parallel projection preserves size, the same object looks exactly the same no matter
how far or near it is. But distance is how our mind recovers the 3rd dimension, depth,
from the 2D image in our retina, so distance is a very important part in visualizing 4D.
It helps us infer 4D depth from a 3D image.
Notice how the same object will appear smaller when it's farther away. By the relative
size of the image, we can infer its distance. This is what we've come to expect in our
own experience in the real world, because our own eyes also see in perspective
projection.
Foreshortening
Perspective projection also has a side-effect: parallel lines in the object are no longer
parallel in the image.
Notice in the above image that parallel edges in the cube are not parallel in the image.
This phenomenon is called foreshortening, and is a consequence of the size of the
image depending on the distance of the object.
However, we 3D beings know that this is really only just a square. It isn't actually
changing its internal angles or the length of its edges. It just appears that way because
of foreshortening in perspective projection as it rotates through 3D space.
Now let's take a look at the analogous situation of a 3D cube rotating in 4D, and see if
we can make sense of it:
The cube appears to be turning itself inside-out and outside-in. One of its faces
appears to be growing and shrinking, and its side faces appear to be distorting into
trapezoids. However, the cube isn't actually being distorted; it only appears that way
because it is rotating through the 4th dimension. More precisely, it is rotating in the
XW plane (the plane defined by the X and W coordinate axes).
We shall explain 4D rotations in more detail later on. But for now, as a help to
understanding this odd effect, consider how the 2D creature would think of the above
animation of a square rotating through 3D. The only rotations it knows are rotations in
the plane, and none of them turn things inside-out like that. Since it has no experience
of 3D rotations, it would perceive the square as distorting and turning inside-out in an
impossible way. But in reality, the square isn't turning inside out; it's just doing a
perfectly normal rotation through 3D. The inside-out effect is merely an artifact of
projecting 3D into 2D.
The envelope of a projected object is the outer boundary of its image. For example,
the following view of the 3D cube has a hexagonal envelope:
The same object may have different envelopes; for example, if a cube is viewed face-
on, it has a square envelope:
A cube can never have a triangular or pentagonal envelope. If you know what
envelopes an object can have, it narrows down what type of object it can be.
However, one should keep in mind that the envelope of an object really only gives
limited information about the object. One should not falsely think that knowing the
envelope of an object's projection is sufficient to uniquely identify the object. For
example, an octahedron also has a square envelope, when viewed vertex-on:
Hence, if you only knew that an object has a square envelope, that doesn't tell you
which object it is. The important thing about a projected image is not the envelope,
but its internal structure. It is the internal structure that gives insight into the structure
of the object. For example, knowing that a cube has a hexagonal envelope isn't very
useful in itself; it is more insightful to see how the 3 projected faces of the cube are
laid out inside that hexagon.
Look at the projection of the 3D cube again. What part of the image do you
automatically focus on? Your attention naturally falls on the central region of the
image, where 3 of the cube's edges meet at the corner that's facing you. In fact, your
attention so spontaneously centers itself on this central region, that you are usually
unaware that this view of the cube has a hexagonal envelope!
But if you were a 2D creature, your viewpoint would be quite different. You would
see this image edge-on, and it would appear as a series of line segments. Upon further
investigation, you would find that these line segments form a hexagonal shape. This is
what your attention would center on. Being told that this image represents a cube, it
would be tempting to identify the hexagonal envelope with the cube. However, the
real point of interest lies inside the envelope.
All this may seem obvious, but it is very important to keep in mind when we start
examining projections of complex 4D objects into 3D. These projections often have
fascinating envelopes, such as rhombic dodecahedra, cuboctahedra, and other
interesting polyhedra. It is tempting to unconsciously identify these polyhedral
envelopes with the 4D object itself, because we are used to identifying 3D objects by
the shape of their surfaces. However, most of the information about the structure of
the 4D object lies inside the envelope.
Knowing this internal structure of the projection also helps us locate the 4 cubical
cells of the hypercube that are currently visible:
These cells appear to be distorted cubes, but they are actually perfectly regular cubes.
They only appear distorted because they are foreshortened by perspective projection.
When a 4D being looks at the hypercube, its attention falls primarily on this layout of
cubical cells in the image, not on the envelope itself. The envelope is only peripheral;
the inside of the image is what is of interest. When examining projections of 4D
objects, we should always focus on its internal structure rather than be distracted by
its envelope.
Note that in the above views of the hypercube, only 4 of its 8 cells are visible. The
reason for this is that the other 4 cells are behind these four, and are therefore
obscured. We shall explain this in detail in the next chapter.
Which of the above images do you more readily identify as a cube? Probably the
image on the right. Why? Because although the left image is more accurate (it
contains the edges at the back of the cube which are omitted from the right image), it
has too much detail, which becomes distracting. Three of the edges in it correspond
with edges on the back side of the cube, which should currently be obscured from our
view.
In this particular case, the confusion is minimal because only a small number of edges
are involved. But compare the following images:
Which of these images help you see the shape of the dodecahedron better? Obviously
the right one, because even though the left image contains all 12 faces of the
dodecahedron and is therefore more accurate, the faces at the back visually clash with
those in front, causing a tangle of lines that obfuscates the whole image. There are
many illusory kite-shaped, diamond-shaped, and tetragonal faces that aren't actually
faces. In fact, it is difficult to see any pentagonal faces at all! The right image, on the
other hand, does not suffer from these problems, because hidden surfaces have been
removed from it.
These examples make it clear why hidden surface removal (HSR) is important in
projections. Although we often think more information is better, sometimes clarity is
more important.
Removing hidden surfaces also helps us tell what is in front and what is behind. For
example, consider the following oblique projection of the 3D cube, also known as
the tumbling cube illusion:
It is hard to say whether it is the lower left face or the upper right face that is the front
of the cube, because this image doesn't tell us what is in front and what is behind. If
we applied hidden surface removal, however, it would remove all doubt:
There is a middle ground between leaving all edges in the image and risking visual
confusion, and completely removing hidden edges, thereby losing some information
about the original object. Instead of completely removing obscured edges, we can
indicate that they are obscured by using visual cues such as lighter lines.
For example, the following image of the dodecahedron uses a lighter color for edges
that are hidden:
Now we can see all the edges and faces of the dodecahedron, and it is less confusing
which ones are in front and which ones are behind.
We can further enhance these images by using thin cylinders for edges and little
spheres for vertices:
We shall now look at how these image enhancement methods may be applied to 4D
projections.
Enhancing a 4D Projection
Now, projections of 4D objects are usually very complex. If we simply apply these
steps without any enhancements, the result will often be very hard to understand.
Sadly, many Java applets and other programs that perform projections of 4D objects
use line-drawings with no enhancements applied. As a result, all we see is a large
mass of tangled lines that leaves us totally confused. We need to apply some
enhancements, at the very least, HSR (hidden surface removal).
However, as we have seen in the previous chapter, the internal structure of the 3D
image produced by the first step is also very important; so we cannot blindly apply
HSR to both steps. In particular, if we apply HSR to the second step, only half of the
projection envelope is left of the 3D image. The result would not be very insightful at
all.
Therefore, we should apply HSR to the first step so that the resulting 3D image is not
cluttered, and then we apply HSC (hidden surface cueing) to the second step, so that
the internal structure of the 3D image is retained, while making it easier to discern the
3D shape in the resulting 2D image.
Confusing, isn't it? Now let's apply HSR by removing edges hidden from the 4D
viewpoint:
Better, but it is still difficult to discern the 3D shape of this projection because we're
projecting the 3D shape into 2D. It is possible to add more enhancements to our line-
drawings to make it easier to understand, but ultimately, the only real solution to the
fundamental limitations of line-drawings is to discard them and move on to 3D
renderings with transparent faces, and lighting and shade:
The 3D structure of this projection is now much clearer. Finally, we add one last
enhancement to this image by using thin cylinders and spheres for edges and vertices:
Now the image is much easier to understand. We can see that it has a rhombic
dodecahedral envelope, with three internal edges that meet at a vertex inside the
envelope. It is similar to the rhombic dodecahedral projection of the hypercube that
we saw in the previous chapter.