EXISTING TECHNIQUES OF VISUAL PROCESSING - Video Object Extraction and Representation: Theory and Applications - page 27

How Do We Best Leverage Current Research?

Video Object Extraction and Representation must use

contributions from many different fields of research.

Research

Contributions Caveats

- A working system
- Insights into

- Different

Implementation

Biological
Vision
Systems

design process (Silicon vs. Cells)

- Insights into visual
processing

- Ambiguous

Design Choices

- Same problem formulation - No temporal

- Well-tested Feature aspect

Image
Segmentation

extraction techniques -Works with

Better Data - Systems that integrate

object/scene knowledge Quality

- Formalizations of
video sequence behavior

- requires many

Motion
Estimation
(ME)

assumptions

- alternative representation
of video sequence

- motion info. is

ambiguous

-a prime example of
content-based processing

- Heavy

Computation

- another source of

- requires

Knowledge
Representation

information for analysis classification

technologies - a technology for search

- optimization technology

- restricted

Dynamic
Programming

- low complexity application

domain
- compression
and content are
not the same

Compressed
Domain

- a source of content
analysis w/o computation

We can use these technologies in our systems,

but we must be mindful of their problems.