Databases Reference
In-Depth Information
Figure 2: A moving camera in a 3D virtual space and the corresponding video replayed in
another window
for all frames of a video sequence as arrow icons or region icons on a 2D map, the
number of arrow or region icons becomes too large to browse and click. We must
simplify such many icons for video sequences. For example, many arrow icons
can be represented by a small number of arrow icons. The representative arrow
icon represents video sequences or a camera's movement for a certain duration. A
video sequence is replayed when the corresponding icon is clicked by a user.
There is a problem of how to divide a video sequence into multiple video
subsequences which are represented by icons. For example, each arrow icon may
represent a segment of every 5-minute video sequence. It is useful to divide a video
sequence into more meaningful sub-sequences, but it will be much more difficult
than dividing them into pieces of constant duration sub-sequences. A simple method
of generating a representative arrow icon for a video sequence is to use the average
values of the positions and directions of a set of time-series momentary cameras.
The method often fails because it is generally difficult that an average value
represents all values for any case. Methods of generating representative arrow
icons should be appropriately selected for various cases.
We can extend the idea of spatial browsing of videos on a 2D map to the one
for a 3D CG space. All 3D CG spaces discussed in this paper correspond to existent
spaces in the real world. Such 3D CG spaces corresponding to parts of the real
world are called
3D virtual spaces
in this paper. We walk through a 3D virtual
space, and browse or retrieve our intended video sequences which look similar to
the current views in the current 3D virtual space. These requests of interactions
