Digital Signal Processing Reference
In-Depth Information
Figure 6.17.
Our results from object measures in comparison to other systems from
MPEG-7 Core Experiments for Shape from MPEG-7 February 1999 Meeting in Lan-
caster, England. See Section 6.14 for results.
Since video objects have no well-defined objective qualities, the MPEG-7
syntax supports many different types of description schemes (DS) that
allow the user to search for video objects efficiently, one of which is based
upon shape. This chapter introduces a shape descriptor for video objects
based on assumptions of implicit support structure as the feature of a
object, i.e., the skeletal structure implied by a object's image boundary.
We use robust representation and comparison of these skeletal structures
as a shape similarity measure via extraction by VOS and representation
by DOTs. This shape descriptor is competitive with the state of the art
technologies and also implements new functionalities such as invariance
to bending at joints and motion annotation. We can use this shape query
system to classify an unknown object's skeleton in MPEG-7 database or
to search MPEG-7 database for video objects with the similar skeleton
structure.
To conclude, Chapter 7 discusses the implications of extraction and
representation system of the work and its future of computing with
content-based information processing.
Search WWH ::
Custom Search