Multimedia Database Management System MDBMS

1
Multimedia Database Management SystemMDBMS

• Presented by
• Mark Stopyro
• Puthyrak Kang

2
Multimedia DBMS

• Large and new topics for research
• Many opening research topics from the data
  model, to storage to efficiency of MDBMS
• Special focus on Content Base Visual Information
  Retrieval.

3
MDBMS

• Presentation focuses on
• General Overview of MDBMS
• Definition
• Characteristics of MDBMS
• Tasks of MDBMS
• MDBMS Design
• MDBMS Architectures
• Content-based Retrieval Mechanism
• Advantages of MDBMS

4
MDBMS Definition

• Multimedia Data
• Michae3 defines as many kinds of mediaimages,
  video, audio, graphics, hypertext, hypermedia,
  and other abstract data types.
• Multimedia Object
• Timothy2 a multimedia document or presentation
  containing one or more multimedia data.

5
Definition (continued)

• Multimedia Database
• Ross Lee Graham10 defines as a database
  containing one or more multimedia object.

6
Multimedia Document
7
Characteristics of MDBMS

• Large object size
• Synchronous delivery of multimedia objects
• Multimedia objects may have embedded timing
  constraints
• Multimedia object composed of multiple components

8
Characteristics (continued)

• Queries are not text or numeric based, but
  Content-based
• Most multimedia transactions are long and
  requires long processing and retrieval time
• Multimedia Object presentation is very important.

9
Main Tasks of MDBMS

• Enforce data independencephysical and logical
  level
• Organize, classify and query multimedia data
• Support abstract data analysis for searching
  engine

10
Tasks of MDBMS (continued)

• Support asynchronous requests and operations
• Provide cost effective storage management scheme
• Ensure data integrity constraints
• Provide basic operations as supported by
  traditional databasesinsert, delete, search and
  update.

11
Tasks of MDMS (Continued)

• MDBMS also has to support
• Composition and decomposition of Multimedia
  objects
• Security and intellectual property protection
• Concurrency control and locking mechanism
• Recovery
• Indexing and clustering

12
MDBMS Design

• Approaches
• File system Not flexible, scalability
• Relational DBMSNo Content-based indexing or
  query support
• OODBMSBinary Large Objects (BLOB)
• Design from the scratchTime consuming.
• ? OODBMS

13
Multimedia OODBMS

• Implementation difficulties (e.g. Concurrency
  control) due to
• Lack of control over OODBMS
• MDBMS requirementsQuality of Service,
  Synchronization, and Networking

14
MDBMS Architecture

• Timothy2 three layers
• Interface
• Object composition
• Storage
• Interface object browsing, query, compose and
  decompose
• Object Composition Manage multimedia objects
• Storage Clustering and indexing

15
MDBMS Architecture continued
16
ArchitectureDistributed MDBMS
17
Content-Based Visual Information Retrieval

• An investigation of current methodologies,
  limitations, and directions for the future

18
Image Retrieval Methodologies

• Free browsing
• Text-based keyword searching
• Contentbased searches

Multimedia retrieval systems often
use Combinations of methodologies
19
Free browsing

• Present user with a set of links to images
• May include summaries (e. g. thumbnail images or
  video key frames)
• Links may be structured (categories, hierarchies)
• Easy to implement Images and/or links may be in
  a database
• Suitable for casual and infrequent use only

20
Text-based Searching

• Descriptive metadata annotations
• Compatible with conventional query models
• Semantically rich
• Considerable human effort required
• Highly subjective annotation process
• Not scalable to large or rapidly expanding
  collections

21
Content-based Searching

• Index binary content - feature extraction
• Machine-generated indexes based on extracted
  features
• Eliminates need for human annotation

22
Content-based Searching (cont.)

• Better scalability than previous methods
• Requires similarity-based query models
• Semantic mapping is the main difficulty
• Semantic mapping is the main difficulty

23
Feature Extraction

• Features can include colors, shapes, textures,
  and motion
• Perform aggregation and dimensional reduction
• Perform similarity and distance calculations
  among derived features
• Indices store calculated values

24
Color Feature Example

• Count occurrences of each color?
• 24-bit color 16.7 million colors
• Aggregation
• Arbitrary (usually even increments)
• Perceptually similar (uneven distribution)
• Relative area of each color range

25
Color Feature Example (cont.)
Image
Aggregation
Histogram
Images courtesy of NASA/GSFC/METI/ERSDAC/JAROS,and
U.S./Japan ASTER Science Team
26
Content Based Querying

• Goal Return the K most relevant results ranked
  in order of relevance, while excluding as much
  non-relevant information as possible
• Challenge results must be semantically relevant
  in the context of the users intent

27
Information Abstraction Levels
Semantic Queries
Syntactic Queries
28
Abstraction Level Querying

• Data find copies, subsets, supersets
• Feature find images with X of colors a, b, c
• Object find images containing red cars
• Concept find images of distressed students
  studying for final exams
• The Semantic Gap is still unbridged

29
Narrowing The Semantic Gap

• Continued refinement of feature extraction and
  indexing techniques
• Increased volume of generated metadata
• Injection of domain knowledge
• Interactive query refinement Relevance feedback

30
Future CBVIR Directions

• Improve relevance feedback mechanisms
• Increased web orientation
• Standards for evaluating CBVIR systems
• High dimensional indexing techniques
• Understand human perception
• Better user interfaces
• Cross discipline cooperation between DB and
  imaging communities

31
Advantage of MDBMS

• Multimedia Application playback and
• production (Script Language).
• Flexibility of script language
• Shared script and data structure
• Reuse of Multimedia objects
• Reduce data redundancy
• Data independences physical and logical
• Improve Search and Retrieval

32
Research Topics

• Issues in multimedia database management systems
  4
• Synchronization and Storage Models for
  Multimedia Objects5
• Database system for video objects6
• Layered multimedia data modeling7
• An object-oriented multimedia database system for
  a news-on-demand application8
• Design of a multimedia object-oriented DBMS9
• Multimedia database management (NSF, Fuji
  Electric, ATT)
• Video modeling and management
• Multimedia document management
• Distributed multimedia systems (NSF, AFRL, IBM,
  Intel, Siemens)
• High-performance multimedia database
  architecture for storage management(NSF, ATT)
• Development of video processing techniques for
  automatic object identification and motion
  tracking
• Development of data models with powerful semantic
  expressiveness

33
REFERENCES

• 1.        Huan-Chao Keh, and Timothy K. Shih,
  Formal Specification in Software Reuse Designs
  an Object-Oriented Database Example, Tamkang
  Journal of Science and Engineering, vol.1, No. 2,
  pp. 97-113 (1998)
• 2.        Timothy K. Shih, Distributed
  Multimedia Database, Multimedia Information
  Network Lab, Department Science and Information
  Engineering, Tamkang University, Taiwan, 2001.
• 3.        Michael M. David, Multimedia
  Databases Through the Looking Glass, Intelligent
  Enterprises Database, 1997
• 4.        Raymon Paul, M. Farrukh Khan, Ashfaq
  Khokhar, and Arif Ghafoor, Issues in Database
  Management of Multimedia Information, in
  proceeding of the 18th IEEE annual international
  computer software and application conference,
  1994
• 5.        Thomas D. C. Little and Arif Ghafoor,
  Synchronization and Storage Models for
  Multimedia Object, IEEE Journal on Selected
  Areas in Communications, Vol. 8, No. 4, April,
  1990.
• 6.        Keh-Feng Lin, Chueh-Wei Chang, and
  Suh-Yin Lee, Design of an Interactive Video
  Database, in proceedings of the 1994 HD-Dedia
  Tech.
• 7.        Gerhard A. Schloss and Michael J.
  Wynblatt, Presentation Layer Primitives for the
  Layered Multimedia Data Model, in Proceeding of
  the IEEE 1995 International Conference on
  Multimedia Computing and Systems, 1995.
• 8.        M. Tamer Ozsu and Duane Szafron and
  Ghada El-Medani and Chiradeep Vittal, An
  object-oriented multimedia database system for a
  news-on-demand application, Multimedia Systems,
  1995.
• 9.        C.Y. Roger Chen and Dikran S.
  Meliksetian and Martin Cheng-Sheng Chang and
  Larry J. Liu, Design of a multimedia
  object-oriented DBMS, Multimedia Systems, 1995.
• 10.    Ross Lee Graham, Multimedia Database
  Technology an introduction to MMDB, ITM
  Mid-Sweden University
• Thomas C. Rakow, Erich J. Neuhold, and Michae
  Lohr, Multimedia Database System The Notions
  and Issues Integrated Publication and
  Information System Institute (IPSI).

34

Reference Literature
12 Marques, O. and B. Fuhrt. 2002.
Content-based visual information Retrieval, in
Distributed multimedia databases techniques
applications. T. K. Shih, editor. Pages 35
57. 13 Chua, T. S., W. C. Low and T. M. C.
Sin. 1998. Visual Information Retrieval. The
Journal of the Institute of Electronic Engineers
of Japan. 27(1) 10 19.
35

• THANKS
• QUESTIONS
• COMMENTS