View
163
Download
0
Category
Tags:
Preview:
Citation preview
EFFECTIVE AND EFFICIENT QUERY PROCESSING FOR
VIDEO SUBSEQUENCE IDENTIFICATION
BY MADHUKAR REDDY 08911A0516RAHUL P 08911A0532
PROJECT IN CHARGE-MR M.RAVI(HOD ,CSE)
Objective
o To find a generic database management solution towards effectively and efficiently searching similar videos, with tolerance to different variations introduced during not only transformation process but post-production editing.
o To retrieve similar frames from two videos
Why do we need this?
o To aid Recognition for Copyright Enforcement
o TV Commercial Detectiono Aid Investigating Agencieso Aid Film Certification Board
Introduction
o Rapid advances in multimedia and network technologies have popularized many applications using video databases
o A video sequence is an ordered set of a large number of frames
o Each frame is represented as a high-dimensional vector
How the existing system works?
o Content Based Video Retrieval o We need to check manually
whether a video is a part of a long stream by browsing its entire length
o Identification cannot be done in case of any editing in the video
o Time taking
Features of Proposed System
o Subsequence based video retrievalo Any subsequence of a long database
video that shares a similar content to a query clip is retrieved
o Relevant videos can be identified even if there exist transformation distortions, partial content reordering ,insertion, deletion or replacement
o Time efficient
Preferred Technologies
o Windows XP Operating Systemo JDK 6.0o Java Media Frameworko Edit Plus
Modules
oVideo Copy Detectiono Sliding of query video frame by frame on
database video with a fixed length windowo Globe signatures have been used to avoid
distortions while video transformationso Video is depicted globally o Applicable for queries with multiple shotso Detects videos of same temporal order and
length
oVideo Similarity Searcho Shortcomings of the former module are rectifiedo Sub sampled frame-based matching is done o Average inter frame similarity is taken into
considerationo Frame alignment, gap, noise for accurate
identification are also consideredo Scores of different factors are aggregated to
derive the most similar subsequence based on overall video similarity
Feasibility Study
o Is it worth doing ?o Technical Feasibilityo Operational Feasibilityo Economic Feasibility
Software Requirement Specification
o Spiral Model o Different stages of SDLCWhy Spiral Model?o Estimates become realistico Easy to cope with changes
Functional Requirements
o Input Videoo Database Videoo Conversion of Input Video into frameso Conversion of Output Video into
frameso Subsequence Identificationo Frames matchingo Show the duplicate frames from the
input video
Data Flow
Convert into No.of Frames
Select Input VideoSelect Database
Video
Sub Sequence IdentificationSub
SequenceYes
Show Subsequence Frames
Content Verification
Yes
Use Case Diagram
User
Select Input Video
Select Database Video
Sub sequence identification
Convert Both Video into No.of Frames
Shows the sub sequence frames
Sequence DiagramUser Input Database Frames sub sequence Duplicates
1 : input video() 2 : convert to frames()
3 : DB video()
4 : convert to frames()
5 : both frames()
6 : compare()
7 : duplicate frames()
Collaboration Diagram
UserInput
Database
Framessub sequence
Duplicates
1 : input video()
2 : convert to frames()
3 : DB video()
4 : convert to frames()
5 : both frames()
6 : compare()
7 : duplicate frames()
Class Diagram
Activity Diagram
Select Input Video Select DB Video
Convert to Frames
Sub sequence identification
Show duplicate frames
Testing
o Correctness, completeness, security and quality are achieved
o Manual testingUnit testingSystem TestingAcceptance TestingRegression Testing
o Operation testing is done by testing whether all components perform its intended operations
Implementation
o Java Swingso JLabels, JFrames,
JTextArea,JList,JFileChoosero Inherited from JComponent classo Pluggable look and feelo Javax.swing package
Snapshots
Future Enhancements
o Improvement of Video clarityo Elimination of distortionso Investigate the effect of representing
videos by other features, such as ordinal signature
o The weight of each factor for measuring video similarity might be adjusted by user feedback to embody the degree of similarity
Conclusion
o Similar frames are retrieved by algorithms
o Bipartite graph is constructedo Dense segments are choseno Irrelevant segments are prunedo Relevant segments are processed
Bibliography
o A.W.M. Smeulders, M. Worring, S. Santini, A. Gupta, and R. Jain, “Content-Based Image Retrieval at the End of the Early Years, ”IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 22, no. 12,pp. 1349-1380, Dec. 2000.
o C. Faloutsos, M. Ranganathan, and Y. Manolopoulos, “Fast Subsequence Matching in Time-Series Databases,” Proc. ACM SIGMOD ’94, pp. 419-429, 1994.
o H. Wang, A. Divakaran, A. Vetro, S.-F. Chang, and H. Sun, “Survey of Compressed-Domain Features Used in Audio-Visual Indexing and Analysis,” J. Visual Comm. and Image Representation, vol. 14, no. 2, pp. 150-183, 2003.
o R. Mohan, “Video Sequence Matching,” Proc. IEEE Int’l Conf. Acoustics, Speech, and Signal Processing (ICASSP ’98), pp. 3697-3700, 1998.
o C. Kim and B. Vasudev, “Spatiotemporal Sequence Matching for Efficient Video Copy Detection,” IEEE Trans. Circuits and Systems for Video Technology, vol. 15, no. 1, pp. 127-132, 2005.
o X.-S. Hua, X. Chen, and H. Zhang, “Robust Video Signature Based on Ordinal Measure,” Proc. IEEE Int’l Conf. Image Processing (ICIP ’04), pp. 685-688, 2004.
Thank you!
Recommended