Conclusion - Real-Time Rendering: Computer Graphics with Control Engineering - page 117

VDR with Occlusion Culling

20

15

10

5

VDR without Occlusion Culling

0

100 200 250

0 50 150 300

Frame number

FIGURE 8.9 Quick-VDR: Interactive View-Dependent Rendering of Massive Models,
Yoon et al. [81].

uses the feedback algorithm. However, it only presents to the reader a single frame
rate objective.
Similarly, the experiment results from Yoon et al's work [81] on interactive
view-dependent rendering of massive models show improved frame rates but only
with oscillations and unstable frame rates. There is no depiction in Figure 8.9 of any
adaptive capability in handling discrete frame rate level changes.
Another example that shows frame rate improvement is found in Figure 8.10 from
Scherzer, Yang, and Mattausch's research [69] on exploiting temporal coherence
in real-time rendering. The output from their technique is compared against other
approaches but there is no information on the technique's handling of frame rate
level transitions and the corresponding transient response.
Our proposed control system for 3D rendering however produces fast and direct
transitions from one steady-state transition to another as shown in Figure 6.12 in
Chapter 6. The small amount of delay in the tracking is due to the implementation of
the controller and plant which involves network communication.

8.2 SUMMARY
In this topic, we described an intelligent real-time rendering system based on our
research in the fields of control engineering, system identification, and real-time
computer graphics. We introduced a novel control system framework using a
closed-loop feedback design with the rendering process—the plant to be controlled.
The salient areas of this research were the detailed process steps for deriving system
models for the real-time rendering process. The techniques were not discussed in
previous research. We devised models that can capture both linear and non-linear
characteristics of the rendering process.