INTRODUCTION TO CONTENT-BASED VISUAL PROCESSING - Video Object Extraction and Representation: Theory and Applications - page 22

called DAG-Ordered Trees. Using the previous work in DAGs
in a multi-resolution framework, we can produce a robust shape
representation in Chapter 6.
4. A Novel Video Object Query By Shape System
Applying an extraction technique called Voronoi Ordered Skele-
ton (based upon Voronoi Ordered Spaces), we create a shape
query system with novel functionality. Using the robust repre-
sentation of DAG-Ordered Trees and the extraction of multireso-
lution representations from Voronoi Order Skeletons of Chapter 3,
we implement our video object shape query system.
8. Topic OUTLINE
The topic is outlined as follows. Chapter 1 places the topic in refer-
ence to various fields of study: Image Segmentation, Computer Vision,
Motion Estimation, Neural Networks and Adaptive Signal Processing.
Chapter 2 builds a working vocabulary of algorithms and issues asso-
ciated with video processing. Chapter 3 describes the key concept of
Voronoi Ordered Space as a means of both describing a priori shape in-
formation and extracting shape information from a given contour. Build-
ing on Chapter 2 and Chapter 3, Chapter 4 presents a system for Video
Object Segmentation. Chapter 5 outlines the general methodology of
representing data with Directed Acyclic Graphs (DAGs). Building upon
Chapter 3 and 5, Chapter 6 presents a system for Video Object Query
by Shape in support of the MPEG-7 standard. Chapter 7 proposes fu-
ture applications of content-based information processing and links our
MPEG-4 and MPEG-7 efforts together, presented in Chapter 4 and 6,
respectively, as a single adaptive system.
9. CONCLUSION
We have placed our research within the context of current technologies
and directions of future multimedia technologies. Content-based video
processing requires a convergence of varied fields of studies to provide
the theoretical framework.
A major theme of this topic will be the integration of high-level infor-
mation into low-level computation. From the range of problems given by
the MPEG standards, we concentrate upon two major problems, video
object extraction and representation that are central to the MPEG-4
and MPEG-7 standards, respectively. From the solutions to these prob-
lems, its supporting algorithmic components, and future systems, we
hope to shed light how systems translate their high-level representations
of knowledge into their low-level extraction operations. Finally, since we