the simulation whilst some modifications are performed.
3.3 MR registration and staging
Employing a markerless camera tracking solution for registering the CG camera according to
the real one, is an added value advantage since it eliminates the use of external tracking devices
or avoids polluting the real scene with the use of known fiducial markers. However, the issue
that arises is how to geometrically calibrate the camera and define the scene fiducial origin in
world coordinates. Especially as our MR scenes have animated virtual characters, initial character
staging, scaling and orientation is a crucial factor in order to determine correct initial, lifesized,
believable geometrical registration. In the pipeline described in section 4, boujou allows
for an initial scene origin to be defined offline on a tracked scene feature. This feature though is
not sufficient as for a number of characters and a storytelling scenario, designers would like to
interactively direct, stage and adjust the action in real-time, according to their dramaturgical interest.
Therefore we propose a simple algorithm for determining the storytelling scene origin
and orientation, harnessing the features of the underlying OpenGL scenegraph renderer camera
metaphor and world node coordinates as depicted in Fig. 3. We allow for interactive manipulation
of the scene camera as well as separate scene global repositioning and scaling according to
the standard OpenGL formulas for ModelView and Projection matrices: The ModelView and
Projection matrices are used to set up the virtual camera metaphor and are provided in real-time
for each tracked frame, by the underlying camera tracker (Papagiannakis et al 2005). Thus for
interactive authoring-staging in the MR scene, two more controls are supplied: a) A single vector
is mapped via keyboard contols on the translation part of the camera so that the camera can
be furthermore tweaked within the tracked frame and b) A single 4×4 translation matrix is
mapped as a virtual trackball metaphor, so that a designer can interactively stage the “Main
Scene” scenegraph node that contains all the virtual augmentation. Since the basic renderer of
the VHD++ framework is based on OpenScenegraph, all individual elements can have their
transformation matrix modified. However, for the MR real-time authoring stage, it is important
that the whole staged experience can be initially positioned according to the real scene so that
the camera tracking module can subsequently register it accordingly with the real scene.
Fig. 3. Camera Coordinate Systems in MR Fig. 4 Stylized Mediated Reality Extension (The “Visual filter”
Welcome to Authspot, the spot for creative writing.
Read some stories and poems, and be sure to subscribe to our feed!