Digital Signal Processing Reference
In-Depth Information
Chapter 4
A SYSTEM FOR
VIDEO OBJECT SEGMENTATION
In this chapter, we implement a video object segmentation system that
integrates Voronoi Ordered Spaces of Chapter 3 with existing techniques
of Chapter 2 to support the MPEG-4 functionality of object-addressable
video content [Lin and Kung, 1998a] [Lin et al., 1998] [Lin and Kung,
1999], Our surface optimization formulation describes the video object
segmentation problem (see Section 6.) in form of an energy function that
integrates many visual processing techniques. By optimizing this sur-
face, we balance visual information against predictions of models with
a priori information and extract video objects from a video sequence.
Since the global optimization of such an energy function is still an open
problem, we decompose our formulation to allow a tractable optimiza-
tion via dynamic programming in an iterative framework. In conclusion,
we show the results of the system on the MPEG-4 test sequences and
objectively compare them to results that are hand-segmented by the
MPEG-4 committee.
1. PREVIOUS WORK
The MPEG-4 standard is the first major step to support content-
based functionality (such as cut and paste, synthesis, and query of video
objects) with its object-addressable content (see Figure 4.1). The major
enabling technology for the MPEG-4 standard are systems that compute
video object segmentation,
i.e., the extraction of video objects from a
given video sequence.
Research in video object segmentation has precursors in object track-
ing and detection technologies [Irani and Peleg, 1992] [Wang et al., 1995].
However, instead of extracting the video objects, object tracking only
derives points on an object that represent the motion of the object as
Search WWH ::
Custom Search