Databases Reference
In-Depth Information
2
An Overview of Spatial Browsing for Video Data
This chapter overviews the use of spatial description data of video data for browsing
and retrieving video data of users' interests using spatial queries. A video data is
considered a collection of time-series images in this paper. First, we discuss the use
of some spatial description data of
photo pictures,
that is, images, then extend it
to the use for videos later. In the remainder of this paper, a picture means a photo
picture which is one image. One picture corresponds to a camera at a certain
moment. The camera at a certain moment is represented by some spatial attributes,
such as its position, direction, zoom ratio and so on. 2D (dimensional) map data
are also used for making their relations to some spatial attribute data of the camera.
If the time-series positions of the camera for some duration are visualized on a 2D
map, the visualization can show the distribution of the camera's movement which
corresponds to a collection of time-series pictures. The position data can also be
used to create clickable icons representing time-series pictures. If we click one of
the position icons, the picture corresponding to the clicked icon will be shown on
a screen. Also, we use the direction and positi on of the camera to represent cameras
or pictures. In this case, the camera can be visualized as an arrow which provides
the information on the direction and position. The arrow on the 2D map enables
users to understand which direction's scene can be viewed in the picture. If
information about the position, direction and zoom ratio of the camera are
available, we can tell what region in the real world was taken or viewed in a
picture at a certain moment. The region in the 2D map can also be used as a
clickable icon for users' interactions. If we click the clickable region icon, the
corresponding picture will be displayed on a screen.
The spatial data, such as position and direction of a camera, can be used for
spatial queries. The spatial data are also used as clickable icons for 2D map
hypermedia as we mentioned before. For example, if we want some pictures of
Mt. Fuji, we can realize it by making a spatial query to find regions containing the
point representing Mt. Fuji. The regions represent pictures. Thus, we can indirectly
retrieve pictures by spatial queries on the contents in the pictures using spatial
data corresponding to the pictures. We can extend this idea to applications of
video data. Video data can be considered a collection of images or pictures. The
unit images comprising a video sequence are called
frames
of the video sequence.
Each frame of the video sequence corresponds to a camera at a certain moment
and it has its spatial description information (Figure 1). Figure 1 shows the
correspondence between a camera's momentary condition and its image as well as
some temporal relations of time-series images. We also make some spatial queries
for video sequences. For example, we can make a query to select some video
sequences which show Mt. Fuji. For cameras taking videos, their time-series
positions and directions are expected to be automatically recorded by using spatial
sensors in the near future. We can see the camera's movement and replay the
corresponding video at the same time. If we visualize all spatial description data
