![](/image/demo/project_title.gif) |
|
![](/image/demo/project_overview.gif) |
|
Our goal is to propose a flexible and efficient solution to
generate summary for sports video data. To this end, we focus
on incorporating the standard audio-visual features (MPEG-7
descriptors) with domain-specific information in a common framework,
and thus locate and extract interesting events. Currently, our
test dataset are soccer videos. |
|
![](/image/demo/index.1.gif) |
Figure
1: Framework of video summarization system |
|
|
For soccer video: |
|
![](/image/demo/dot1.gif) |
Feature extraction:
two features |
![](/image/demo/dot5.gif) |
Ration of grass color
region that is made from MPEG-7 Scalable color descriptors. |
![](/image/demo/dot5.gif) |
Edge histogram made from MPEG-7 Scalable
color descriptors. |
|
|
![](/image/demo/dot1.gif) |
Shot detection: |
![](/image/demo/dot5.gif) |
three categories of shots:
global view, medium view and close-up view. |
|
|
|
|
![](/image/demo/dot1.gif) |
Scene construction: |
![](/image/demo/dot5.gif) |
two kinds of events: "goal",
and "foul and missing". |
|
|
|
Advantages |
![](/image/demo/dot5.gif) |
The system is flexible
and easy to extend to other domain. |
![](/image/demo/dot5.gif) |
It has high accuracy,
the achieved results are 90.7% of of shot detection,
67%, and 90% of goal, and "foul and missing
shot" event detection, respectively. |
|
|
|
|
|
![](/image/demo/project_publication.gif) |
|
|
Nguyen Ngoc Thanh, Truong
Cong Thang, Beet Nara Bea, Yong Man Ro, "Video Abstraction
Generation Supervised by User Preference", SPIE 2003 |
|
Nguyen Ngoc Thanh, Truong
Cong Thang, Tae Meon Bae, Yong Man Ro, "Soccer Video Summarization
System Based on Hidden Markov Model with Multiple MPEG-7 Descriptors",
Submited to CISST 2003 |
|
|
![](/image/demo/project_demo.gif) |
|
Click here
to see examples of soccer video summary. |
|
|
|