Upload
mkoskela
View
614
Download
1
Tags:
Embed Size (px)
DESCRIPTION
Talk at VIPP 2010 Symposium, Jan 12th 2010.
Citation preview
VIPP Symposium, January 12, 2010
Content-Based Video Analysis
Markus Koskela
Background
Aalto University School of Science and Technology
Department of Information and Computer Science
(http://ics.hut.fi/ )
Adaptive Informatics Research Centre, Academy CoE(http://www.cis.hut.fi/research/ )
Erkki Jorma Ville Mats He Xi JingOja Laaksonen Viitaniemi Sjoberg Zhang Chen Wu
(2/20)
Overview of my talk
• Introduction
• Concept-based video search
• Video summarization
• Video analysis in AR applications
(3/20)
Content-based analysis of video
• Visual features
• Audio features,
speech, text, . . .
(4/20)
Indexing with low-level features
(5/20)
Content-based image retrieval
(6/20)
Google Images
(7/20)
Types of video material
Shot-based video, obtained with
shot boundary detection
Continuous material, sampled
with regular interval
Online video, received one frame
at a time
(8/20)
Shot-based video indexing
Texture SOM
Temporal Color SOM
Motion SOM
Word histogram Keyword freq.Color SOM
text
video shot
launch...
...missile
keyframe audio
Audio SOMbinary keyword
inverted file
(9/20)
Concept-based video search
(10/20)
(11/20)
Concept ontologies
sports weather court office meeting studio outdoor building desert
snowvegetation mountain road sky urban crowd face
US flagperson police/ military prisoner animal computer/TV airplane car
bus truck boat/ship walking/ people explosion/
waterscape/
natural maps charts
waterfront
running
security screen
marching fire disaster
(12/20)
Concept-based video search
(13/20)
Video summarization
(14/20)
Summarization pipeline
(15/20)
Video analysis in AR applications
(16/20)
Contextual IR in AR applications
(17/20)
Devices for mobile AR
(18/20)
Video analysis for AR
(19/20)
Thank you!
(20/20)