Graphics Reference
In-Depth Information
to make simplifications because of the computational complexity. A sometimes-
overlooked important third stage (and one we do not consider at all in this topic) is
mapping these values correctly to the display device to produce the correct image for
the user. The goal is to make a user's perceptual reaction to a real scene and to its
picture the same. These comments lead us to say a few words about another topic,
image-based rendering.
Image-based rendering
has become a very active area of research. In very broad
terms, what one is trying to do here is to reconstruct a three-dimensional world from
two-dimensional data that may have been collected from a variety of sensors, such as
cameras. In the case of pictures from a camera, one is starting with something that
is as realistic as it can get since it involves an image of an actual scene, not an artifi-
cially created one. In one form of image-based rendering one takes one or more pic-
tures of a scene and then attempts to show how the scene would look from different
view points. No three-dimensional models are built and no lighting model is involved
because one is simply transforming given pixel values. In other types of image ren-
dering the goal is to reconstruct a three-dimensional world that can then be displayed
from different views. Depending on the generality of the scenes that one wants to
handle, this might involve using tools from the field of computer vision.
Computer
vision
deals with taking some raw two-dimensional pixel data and trying to make
sense of it by, for example, finding the edges in the picture and then collecting them
together into higher-level objects such as curves and rectangles. Knowing the rectan-
gles in a picture one can then try to figure out the three-dimensional object that pro-
jected on them.
By in large, image-based rendering is only peripherally related to the subject of
this topic. A good place for a vast amount of information about the field are some of
the courses at the yearly ACM SIGGRAPH conference that have been dedicated to this
subject, for example, [Debe99]. As one can see from the title of that particular course,
the field can be considered as divided into three parts: image-based modeling, image-
based rendering, and image-based lighting. In
image-based modeling
all or part of the
goal is to reconstruct a three-dimensional world. Although one uses methods of geo-
metric modeling, this is not really a part of the field of geometric modeling per se.
Image-based rendering, as we described above, deals with generating different views
of a picture by transformations of the original.
Image-based lighting
refers to the
process of reconstructing the correct lighting for the different views generated via
image-based modeling or rendering.
10.2
Ray Tracing
10.2.1
A Ray-Tracing Program
A top-level abstract ray-tracing program looks like the following:
