Graphics Reference
In-Depth Information
Image
Vertices
~5500
~2880
~1580
~670
140
Maximum detail,
for closeups.
Minimum detail,
very far objects.
Notes
FIGURE 3.1
Visual effect of varying vertex count for 3D object in discrete steps. (Source:
http://en.wikipedia.org/wiki/Level_of_detail#A_discrete_LOD_example)
experience offered by a real-time rendering application. While the quality of the
generated imagery may be important to the user, the interactive experience is usu-
ally dominated by the application response rather than the quality of the generated
imagery. Furthermore, quality is a subjective notion that complicates the adequacy
of any useful metric.
3.4 LINEAR SYSTEM MODEL REPRESENTATION
FOR REAL-TIME RENDERING
This section describes the modelling process applied to the real-time rendering
system and the derivation of the mathematical models for various types of rendering
applications. Using the system identification methodology, we demonstrate that lin-
ear time-invariant models can be obtained from the input and output data collected
from experiments conducted using sample rendering applications.
A basic relationship between the input and output of a system may be expressed
as a linear difference equation as follows.
()
+
(
)
+…
(
)
=
(
)
+…+
(
)
1 ()
(3.1)
yt
ay t
−
1
a
yt n
−
ut n
−
b ut n
−
−
n
b
+
e t
1
n
a
1
k
n
k
a
b
where:
a
1
…
a
n
a
and
b
1
…
b
n
b
are parameters to be estimated.
y
(
t
) is the output of the system at time
t.
yt
−
(
1
and
yt n
a
(
)
are the previous outputs on which the current output depends.
−
(
)
and
ut nn
(
1
are the previous inputs on which the current output
ut n
k
−
−−+
k
b
depends.
n
a
is the
number of poles of the system or the order of the system.
n
b
represents the
number of zeroes plus one.
n
k
denotes delay in the system.
e
(
t
) equals noise.
An alternative way to represent Equation (3.1) in a more compact manner is the ARX
model described below:
()
()
=
()
(
)
+
()
Aq yt
Bqut net
k
−
(3.2)
Search WWH ::
Custom Search