\newcommand{\tab}{\hspace{2mm}}
\chapter{Overview}


 The visibility work package aims to develop methods for making use of
restricted visibility in large scenes. The major goal of these methods
is to avoid wasting of the computational power on currently invisible
parts of the scene. To address this goal we develop new methods for
online visibility culling and offline visibility preprocessing. The
online visibility culling is either used to render only visible part
of the scene or to provide queries which determine visible part of the
scene in runtime. Visibility preprocessing provides partitioning of the
view space into view cells and for each view cell it calculates a
potentially visible set.


\section{Module Class Hierarchy}

 The visibility work package consist of a module which is integrated
into the game engine and an external module which deals with
preprocessing. One of the goals of the design of the module class
hierarchy has been easy portability of the developed methods to other
game engines. The resulting class structure clearly separates the
actual engine dependent parts from the game engine independent
algorithmic parts.  We have used name spaces to separate the engine
dependent parts (name space Ogre) from the engine independent parts
(name space GtpVisibility). The external visibility preprocessing module
defines its own name space (GtpVisibilityPreprocessor).

 An overview of the important classes of the visibility work package and
their integration into the Ogre engine is depicted in
Figure~\ref{fig:classes}.

\begin{figure}[htb]
  \centerline{
    \includegraphics[angle=270,width=1.0\textwidth]{../classes}
    }
  \caption{Overview of the visibility work packages classes and their integration
    into the Ogre game engine.}
  \label{fig:classes}
\end{figure}

The diagram shows that the module integrated into the engine consists
of three main parts (CullingManager, QueryManager,
PreprocessingManager) which are encapsulated in a helper class
(Visibility Manager). The helper class then deals with the
initialization of the contained parts and allows an easy communication
between them.

\section{Visibility Culling}

Visibility culling seamlessly integrates into the rendering
loop and eliminates most invisible objects from being send to the
pipeline. Visibility culling is implemented by instances of the CullingManager
class. We will provide several implementations which can be easily
switched at runtime and so the best technique for the particular scene
and scene representation can be selected. Note that the Ogre
integration design allows to exploit all already existing methods for
scene management, such as octree, BSP tree, or plain scene graph.


\section{Visibility Queries}

Visibility queries determine the visible geometry for a given view
point in the scene. Additionally they can also report the visible
scene hierarchy nodes. Visibility queries are implemented by
instances of the QueryManager class. Note that some of these instances
will make use of the preprocessed visibility data through the use of
PreprocessingManager.

\section{Visibility Preprocessing}

Visibility preprocessing precalculates visibility for all viewpoints
corresponding to cells of a view space partitioning. The visibility
preprocessor is implemented as a standalone module which imports files
generated by the PreprocessingManager. The results of the visibility
computation will be exported to the file which can then be loaded by
the PreprocessingManager and used inside the engine.

 Apart from the scene definition file the preprocessor will be able
to import the view cell definition. In this case it will assume the
view cells have to be described as meshes satisfying set of
requirements. Alternatively the module can generate view cells by
automatic view space partitioning.


%{\textbullet}\tab\textbf{Construction module for tree multiresolution models.}