Posted by Jeff on January 31, 2001 at 15:04:43:
Here is an interesting article that Alan Reed, a principal at Imagetects, wrote. Thought some of you might be interested.
Using 2D in a 3D World
Getting the most impact from your rendering, with the least amount
With the proliferation of computer graphics technology in
professional practice, there is a great deal of confusion
surrounding how to effectively and efficiently use "2D" or "3D" in
renderings and other visual presentations.
The question we hear most often is, "Are your image libraries 2D or
3D?" Professionals and artists alike feel a push to generate
three-dimensional computer renderings to provide state-of-the-art
presentations to their clients and remain competitive with their
peers. Naturally, when attempting to represent the
three-dimensional world we live in, one would assume that one needs
to develop renderings using three-dimensional objects. This, as we
shall learn, is not true at all.
3D Expressed via 2D
For centuries, we have communicated three-dimensional concepts via a
two-dimensional, flat media -- paper. These renderings employ the
principles of perspective: foreshortening, diminution of size,
overlapping objects and converging lines. These principles have
been well known since the Renaissance. Even computer graphics that
are developed using true three-dimensional objects are eventually
reduced to a two-dimensional view, either to the computer's monitor
or to film, prints, etc.
Even in the world of virtual reality, where the sense of a
three-dimensional environment is maintained, you find the entire
experience built upon 2D images. The reason why 2D imagery is used
becomes obvious: a) a far greater sense of realism, b) simplicity
in development, and c) rendering speed. This first point is perhaps
the most important. The entire purpose of a rendering or simulation
depicting the real world is that it should be an accurate portrayal
of the real world.
Man-made vs. Nature-made
Any real world graphic can be reduced to two aspects, each
complimenting and working off each other. These two aspects are, in
fact, the essence of architecture. They are: man-made geometric
objects and nature-made organic objects. Examples of man-made
objects are planes, cones, spheres, columns, etc. Looking at any
built environment, one can readily see these shapes repeated over
and over again. One thing also becomes clear, all of these objects
have a "surface". While it is true that these objects have depth,
people don't experience this cross-section. Only the surface
characteristics are of any real importance. To accurately represent
these surfaces, 2D surface maps are used. They are called "texture
maps", or sometimes "bitmaps", and are usually developed to repeat
themselves seamlessly over a surface.
Most rendering software is built upon the notion of geometric
space. Yet, a problem arises when we try to extend this concept of
man-made geometric objects into the realm of nature-made organic
People and plants are the nature-made organic objects that bring
life to your renderings and virtual worlds. Capturing this essence
of life eludes mathematicians, no matter how complex their
geometries or fractal algorithms. It is entirely for this reason
that organic objects that populate a scene should be developed from
the real world: capturing the life essence. Not doing so creates a
cold, lifeless, foreign and alien-like world, which certainly is not
the goal of any architect or artist trying to sell the client on a
Speed and Simplicity
The other reasons for the efficient use of 2D organic objects is
rendering speed and simplicity in development. 3D, VRML or virtual
reality models are built entirely from simple 3D objects that have a
2D "surface" texture map applied to them. The 2D organic objects
can be placed in simple rectangular polygons. As will be
demonstrated below, the area surrounding these objects is made
transparent so that the geometric border surrounding the 2D object
disappears, leaving only the object as it would appear naturally.
This technique works equally well for static renderings or dynamic
walk-through's. A "static" rendering is one where, while the
computer model was developed as a three-dimensional construction,
the rendered image is displayed from one set viewpoint. This
viewpoint is what we convey when discussing an architectural or
artists rendering and is used in at least 80% to 95% of all
real-world presentations. Often, these renderings are set up with
light sources to enhance the realism through the casting of shadows.
It is here that 2D photorealistic objects really outshine their 3D
counterparts: for simplicity, speed and life-like realism!
How to Do It
For each organic object, a tree for example, simply place a
proportionately sized polygon in the scene perpendicular to the
"camera" viewpoint. Assign the appropriate 2D image file to the
polygon as a "map" or "decal". In order to assure the proper shadow
angles, it may be necessary to first "flip" this image left to right
and save it with a new file name. This is easily done using any
paint or image processing software. If lighting is being used and
you want to cast a realistic shadow, add a second polygon of similar
size, bisecting the first one. Set the second polygon perpendicular
to the light source, then map the same image file to this polygon.
When rendering in this manner, you have reduced your "face count"
from the potentially thousands it would take to simulate a tree
geometrically (a fake looking one at that) to only two faces. The
result is a real looking scene that retains the life-like essence so
necessary for real world simulation. Repeat this same procedure for
Dynamic 3D, VRML or virtual reality walk-through's operate in much
the same manner as above. By "linking" the polygons to the camera
and light source, as you move from point to point, the object always
appears to be facing you and doesn't become flat. Similarly, with a
moving light source over time (like the sun), linking the object
will maintain a realistic shadow pattern. If the animation portion
of your rendering software is unable to link and pivot objects in
this fashion, another technique works equally well.
Place two or three bisecting polygons at 60 or 90 degree angles to
one another and map each with the same image. In this way, when
moving around the object, it never appears flat, as the next image
fills in for the last, always giving the appearance of a full
object. The perceptual realism to the eye and the users mind is
maintained, overcoming any criticism by those claiming the image
isn't truly three-dimensional.
Using 2D seamless texture maps and real world imagery, specifically
developed for enhancing 3D computer models in the manner outlined
above, is superior to any mathematically generated or geometrically
created object. The advantage of simplicity of model development,
rendering speed and most importantly life-like essence and realism,
demonstrates some of the many benefits gained with using 2D imagery
in a 3D world.