INTRODUCTION TO CONTENT-BASED VISUAL PROCESSING - Video Object Extraction and Representation: Theory and Applications - page 3

3. and have results that are useful to future content-based information
processing.
Instead of treating content-based information processing as a mono-
lithic problem, we break the task as a series of more manageable prob-
lems, transformations from one intermediate representation to the next.
A solution for one of the transformations gives the user a tool to ex-
plicitly manipulate the content and works toward the completion of the
whole content-based information processing task. In this topic, we wish
to duplicate our own visual biological preprocessing in explicit terms for
video processing. In visual processing, we study the intermediate repre-
sentations for both 1) an object-oriented extraction process as technolo-
gies for content-based analysis and 2) an intuitive similarity criterion of
shapes.
2. SCOPE OF Topic
Concentrating on video processing, we now define the scope of this
topic work, in both theory and practice.
From a theoretical perspective, we only deal with the visual projec-
tions that we see rather than the physical object itself. For instance,
in our video object extraction algorithm, we are only concerned with
extracting the areas of the single frame of video onto which a physi-
cal object projects, without considering its physical parameters such as
its 3-D structure or size. We do not cover physical object reconstruc-
tion in this topic and we only cover a limited subset of computer vision
problems. Although some problems that deal with visual projection of
objects cannot be solved unless we consider the physical object itself,
these limitations will be clearly defined and serve to distinguish between
levels of content-based processing.
From a pragmatic perspective, this topic is looking for robust tools
for extraction, comparison, and manipulation of video objects. We con-
centrate on two key functionalities: extraction and representation. In
both cases, since the circumstances and user demands can be so varied,
we must commit to a limited subset of functionality in order to build
complete systems. In the future, we envision a framework of tools to
meet user demands. Although we focus upon specific systems, we can
learn much of the general design methodology for these content-based
processing systems.
3. PHYSICAL OBJECT VS. VIDEO OBJECT
To clarify the objectives of this topic, we define and differentiate two
key terms: the physical object and the video object. The concept of a