Transcript Technical University of Crete Department of Electronic and
Technical University of Crete
Department of Electronic and Computer Engineering E.G.M. Petrakis Multimedia Data Management Euripides G.M. Petrakis http://www.intelligence.tuc.gr/~petrakis http://courses.ced.tuc.gr
Chania 2010 Introduction 1
Definition
Multimedia
: composite entities combining text, audio, images, video (bit-stream objects), graphics
Multimedia Information Systems
: database systems that support all multimedia data types and handle very large volumes of information E.G.M. Petrakis Introduction 2
TEXT
The most common type of information The least space intensive data type The form in which text is stored varies (plain ascii, word files, spreadsheets, annotations, database fields etc.) Text fonts are becoming complex allowing special effects (color, shade, fill etc.) E.G.M. Petrakis Introduction 3
AUDIO
Mbytes), presented as analog, digital or MIDI Analog waveform : electrical signal, amplitude specifies the loudness of the sound speakers Digital waveform audio : digital, less sensitivity to noise and distortion involves larger processing and storage capacities Digital Audio Tape (DAT), Compact discs (CD) WAV (Microsoft’s wave file format) E.G.M. Petrakis Introduction 4
MIDI (Musical Instrument Digital Interface) Furht et.al.96
Commands that describe how the music should be played are stored (instead of sound) A music synthesizer generates sound Provides high data compression, Widely accepted E.G.M. Petrakis Introduction 5
IMAGES
Digital images: sequences of pixels
Pixels:
numbers interpreted to display intensity, color, contrast etc Binary (0-1 values), gray-scale (8 bits/pixel), colour (3x8 values for RGB) Space overhead depends on image type, resolution, compression scheme
Image formats
: tiff, bmb, jpeg etc. E.G.M. Petrakis Introduction 6
Image Concepts and Structures
Binary images
: 1 bit/pixel black & white photos, facsimile images
Computer Graphics
: 4 bits/pixel
Grayscale images
: 8 bits/sample
Color images
: 16, 24 bits/pixel E.G.M. Petrakis Introduction 7
RGB Representation
A color is produced by adding red, green and blue The straight line R=G=B specifies gray values ranging from black to white E.G.M. Petrakis Introduction 8
YUV Representation
YUV describes the luminance and chrominance components of an image 1 luminance : gray scale version of an image Y = 0.299R + 0.587G + 0.114B
2 chrominance components: U = 0.564(B - Y) V = 0.713(R - Y) E.G.M. Petrakis Introduction 9
Conversions
Conversion
between RGB and YUV requires multiplication operations an approximation: Y = R/4 + G/2 +B/2, U=(B-Y)/2, V=(R-Y)/2 R = Y + 2V, G = Y – (U + V), B = Y + 2U
YCbCr
is another color format for compression Cb = U/2 + 0.5, Cr = V/1.6 + 0.5
E.G.M. Petrakis Introduction 10
VIDEO
The most space intensive data type A sequence of frames Realistic video playback, transmission, compression/decompression require transfer rates about 30frames/sec Microsoft’s AVI and Apple’s Quicktime file formats integrate video and audio in the same presentation E.G.M. Petrakis Introduction 11
Audio-video Modes of Operation
Can be either stored
or
used / transmitted live in real-time Can be used interactively or non interactively E.G.M. Petrakis Introduction Furht et.al.96
12
GRAPHICS
Objects
described through their basic elements (e.g., 2D, 3D shapes) these elements can have different sizes, position, orientation, surface, fill etc. compact representations generated and can be manipulated by design tools (e.g., CAD tools) Their descriptions are stored in files E.G.M. Petrakis Introduction 13
2D and 3D graphics objects
Khoshafian Baker 96 E.G.M. Petrakis Introduction 14
MULTIMEDIA objects
Text, audio, images, video, graphics are elements of complex multimedia objects Various tools or applications integrate, process and combine multimedia Applications : multimedia authoring applications that output documents and databases and end-user applications (e.g., video on demand) Tools : for viewing, updating, querying (presentation viewers, browsers etc.) E.G.M. Petrakis Introduction 15
Multimedia Databases (MDB)
Means stored information or database management systems (dbms)
Multimedia dbms
(mdbms) integrate conventional database capabilities together with different technologies such as Hierarchical storage management (HSM) and Information retrieval (IR) E.G.M. Petrakis Introduction 16
Multimedia Technologies
Technologies integrated within a mdbms HSM IR support support (exact and approximate) Spatial data types and queries Interactive querying, relevance feedback, refining Automatic feature extraction Automatic content retrieval and indexing Query optimization E.G.M. Petrakis Introduction 17
Database capabilities
Persistence : object persist through invocations Transactions : content is inserted, deleted, updated Concurrency control: transactions run concurrently Recovery : failed transactions are not propagated to the db Querying : content can be retrieved Versioning : access previous states of objects Integrity : transactions guarantee consistency of content Security : constraints for accessing/updating objects Performance : optimal data structures and programs E.G.M. Petrakis Introduction 18
Hierarchical Storage Management (HSM)
Support storage of multimedia objects On-line : on RAM, magnetic disk Near-on-line : on optical storage Off-line : on tapes, shelves Each level has different Performance : decreases from top to bottom Capacity : increases from top to bottom Cost : decreases from top to bottom E.G.M. Petrakis Introduction 19
Information Retrieval (IR) Capabilities
Retrieval is the most common operation Deletions and updates are less common Exact match : search based on exact information Inexact: search based on inexact information e.g., probabilistic The results are ranked by order of relevance Query refinement Iterate over query results Adjust weights of query terms or features And finally resubmit queries E.G.M. Petrakis Introduction 20
MDBMS architecture
Khoshafian Baker 96 E.G.M. Petrakis Introduction 21
MDBMS Implementation
Relies on 3 rd component separate party vendors for each Relational dbms Text retrieval for typical records optical storage module Audio/image/video retrieval Feature extraction system Multimedia object interface for text/audio/graphics/images/video system (e.g. Lucene ) system system E.G.M. Petrakis Introduction 22
Object-Oriented Multimedia Databases
Better design, better suited for multimedia applications Uniform handling of data and operations Data types are objects with internal structures and operations that capture the behavior of objects (e.g., audio playback, video browsing) OO dbms does not satisfy all MM requirements Provides primitives for object handling Multimedia components need to be implemented or integrated E.G.M. Petrakis Introduction 23
Multimedia Applications
Multimedia Systems suggest a variety of applications Multimedia conferencing Multimedia on demand news on demand) (interactive TV, See next page for more … E.G.M. Petrakis Introduction 24
Multimedia Applications
Furht et.al.96
E.G.M. Petrakis Introduction 25
Multimedia Conferencing (MC)
Multimedia conferencing enable a number of participants to exchange multimedia information Each participant has a workstation linked to other workstations over high-speed networks Each participant can send or receive mm data and perform certain collaborative activities E.G.M. Petrakis Introduction 26
A video conference system
Furht et.al.96
The biggest performance challenge occurs when the participants transmit voice and video These are mixed together to form a composite stream consisting of video and voice streams E.G.M. Petrakis Introduction 27
Software Architecture
Furht et.al.96
E.G.M. Petrakis Introduction 28
Architectures
Fully distributed : direct connections between the participants Processing and mixing of media at every location Shortest delay The connections increase rapidly Centralized (star) network : a central is connected to every participant Processing and mixing at central node The central node waits until all media is received before mixing and broadcasting E.G.M. Petrakis Introduction 29
Architectures (cont.)
Double star network : a central node from one star network is connected to another central node of another star network Hierarchical network hierarchically : intermediate nodes, root and leaves (participants) connected intermediate nodes perform mixing and processing the completely mixed data is sent to root who broadcasts directly to the leaves reduces network traffic significantly E.G.M. Petrakis Introduction 30
Multimedia conferencing network architectures
Furht et.al.96
E.G.M. Petrakis Introduction 31
Video on Demand (VoD)
Fast networks coupled with powerful computers and compression techniques will be capable of delivering stream data in real-time On-demand multimedia services interactive entertainment video news distribution video rental services digital multimedia libraries E.G.M. Petrakis Introduction 32
Interactive Television (ITV)
An ITV system must be capable of providing basic TV subscription TV pay per view video on demand shopping education electronic newspaper financial transactions single-user and multi-user games E.G.M. Petrakis Introduction 33
This Course
Emphasis on Text, images, video Information retrieval & systems Data organization Web information systems Semantic Web Video & MPEG standards No emphasis on Architectures Specific applications (VoD, ITV,MC) Services E.G.M. Petrakis Introduction 34