Databases Reference
In-Depth Information
4.3
Classification of video data
We can define the following categories for the types of video data using the above
mentioned time axes.
1.
Material type:
Video data in this category have the same sequences of time
intervals on the media-, record-, semantics-, and logical-time axes. That is,
they should be raw videos just recorded, are not edited, and consist of one
scene.
2.
Relay type:
Video data in this category have almost the same sequences of
time intervals in the record-, semantics-, and logical-time axes. A relay
broadcast of satellite video conference is a typical example. For videos in
this category, there may be gaps on record-time axis. However, There are
no alternation in the sequences of time intervals on the media and record-
time axes.
3.
Cinema type:
The media-, semantics-, and record-time axes of video data
in this category are independent. Most of cinemas and dramas are
categorized into this type.
4.4
Joint operation on logical-time axes
Utilization of logical-time axis enable us to retrieve video objects from several
videos by interrelating separated video segments in terms of logical-time axes as
described above. However, suppose we have several videos of a satellite video
conferencing and a sightseeing in a video database. If there is no causal relationships
between the conference and the sightseeing, then their logical-time axes do not
overlap. As a consequence, we cannot apply joint operation.
In fact, it is a quite complicated problem how to represent logical-time axis.
Conceptually, it is a natural way to represent logical-time axis as the combination
of semantic-time axis and identifiers representing events which are related to the
video, such as “conference” or “sightseeing.” In databases, they can be represented
as a set of time intervals on the semantic-time axis, the serial number of the video
data with the name of the relation representing the identifier of the video data. In
this way, however, we cannot express the video data only with time intervals.
In this chapter, we employ satellite video conference as an example for
observation and experiments. For this reason, we do not need to consider logical-
time axes, since we can consider semantic- and logical-time axes are identical in
this type of video data, as described above.
Let us consider the queries stated in Section 4.1. We can process the first query
by the following operation:
1.
Extract time intervals from one video data's index that includes Mr. X's
appearance and his voice, and apply joint operation on them. Perform the
same operation to the other video data for Mr. Y.
