24-11-2010, 01:56 PM
[attachment=7250]
This article is presented by:Malaya Kumar Das
A VIDEO RETRIEVAL SYSTEM BASED ON COMPRESSED DATA FROM MPEG FILES
WHAT IT DOES?
The system uses a two-stage indexing and retrieval approach :
first to classify video shots into separate databases using their motion characteristics.
second to search each database for colour features.
The two-stage approach reduces the search time to almost half at a slight reduction in retrieval rate.
COMPONENTS IN VIRS
Shot Detection
Shot Representation
Colour Feature Representation
Motion Feature Representation
VIDEO DATABASE POPULATION
Motion-Based Classifier
Four video categories (or types): Sport, News, Drama, and Scenery.
Colour Database
Colour is quantised into 10 colors.
In VIRS each r-frame is partitioned into nine equal rectangular regions: N, NW, W, SW, S, SE, E, NE and Center.
VIDEO DATABASE QUERY
The graphical query interface allows a user to make a query in two steps:
a type query.
a colour sketch query.
The user specifies the type of video shot he or she wishes to retrieve.
The user then enters the colour-based information about the r-frame to be searched from the database of the selected video type.
VIDEO DATABASE QUERY
The graphical query interface consists of:
A visual colour chart of 10 quantised colors with known RGB vectors.
A visual shape chart of common shapes.
A visual texture chart of common textures with known feature vectors.
A menu of video type buttons.
CONCLUSION
This two-step retrieval strategy reduces the search time to half on average at a slight reduction in retrieval rate. The system returns a ranked list of three or more candidates. The user can click on the returned r-frame to play the retrieved video shot and hence, the video clip.