Bilkent University
Department of Computer Engineering

---

S E M I N A R

 

Automatic Detection of Salient Objects for a Video Database System

 

Tarkan Sevilmis

MSc.Student
Computer Engineering
Bilkent University

BilVideo is a video database system for answering spatio-temporal and semantic queries. This system is useful for organizing and querying multimedia data that is time dependent (especially video), which includes but not limited to television broadcast archives. BilVideo handles spatio-temporal queries using a knowledge-base, which consists of a fact-base and a comprehensive set of rules, while the queries on semantic features are handled by an object-relational database. Currently, the processing of videos for spatio-temporal queries is done using tools named Fact-Extractor, which is used to extract spatio-temporal relations between video objects and store them in the knowledge-base as facts, and Object-Extractor, which is used to extract salient objects from video key frames with their user specified Minimum Bounding Rectangles (MBR). These tools operate manually, which is a serious drawback on usability of BilVideo on large datasets. Our task is to automate these tools using computer vision techniques and algorithms such as consecutive frame difference, background subtraction, focus detection, and various other segmentation methods.

 

DATE: April 25, 2005, Monday @ 15:40
PLACE: EA 409