You are on page 1of 34

Manipulating Multimedia Data

• Design application, learning applications and journalist’s


applications are more likely candidates for needing to
manipulate the actual content of media data.
• Multiple users should be able to edit shared data
simultaneously along the lines.
• Any lock should be imposed on as small a part of the data
as possible because, in addition to the problem of
concurrency, the multimedia has the problem of real-time
synchronization.
• Manipulating different kinds of media can be very
computationally intensive and time consuming.
Eg: video search – result is series of video clips presented
and ranked in order.
Single media Multimedia
simple query, using other Composition using several
Single attributes or meta data. different media combining
User images, may need
synchronization, journalist
or medical applications

Concurrent use of same data Users can edit same object


Multi- – real-time broadcasting, simultaneously.
User e.g. Video on demand E.g; Computer Supported
Cooperative Work (CSCW)
Presenting Results
• GUI’s for multimedia – difficult to design
- suffer from high costs
• The main issues that must be addressed are :
- media composition of the result set.
- result set will consist of items that are similar
to the target rather than exactly matching.
- way the result will be ranked in presentation
TM and AM :
Task Application Deals with the context
Model Model - the user’s task model
- the application model itself
Semantic Level

DM: Deals with the dialog model which


Dialog Model describes the low level input activities
that may be performed by the
end-user when using the interface.
Syntactic Level
PM: constructs which will appear on
end user’s display and the
Presentational Interaction Application dependencies between them.
Model Model Wrappers IM: set of possible low-level interactions
b/w the user and system environment.
Lexical Level Concerned with efficiency of interaction
and should be modeled in a way which
is independent of the dialog model.
AW : represents the interaction b/w higher levels and the application functionality.
Includes method signatures and data item declaration.
Text Text Video Screen
Screen Window
Video
Image Window
Image
Window
Screen

Audio Stream
 Multimedia introduces different kinds of relationships between data

 Relationship between data items may be both spatial and temporal

Temporal
 When an object should be presented
 How long an object is presented
 How one object presentation relates to others (audio with video
Types of Multimedia Querying
• Query the content of a media object
• Query by Example
(“get me a video clip of this song”)
• Time-indexed Queries
(“show me a clip 30 minutes after the start of a movie”)
• Spatial Queries
(“ show me a video clip where John is Standing behind
Paul”)
• Application Specific Queries
(“show me where a cell becomes cancerous”)
Issues Of Designing Multimedia Database
Systems
• Capturing relationships that involve
i) time
ii) duration
iii) synchronization
• One of the most challenging media to model is video.
• A suitable model could include references to
-- physical objects (people, buildings, vehicles)
-- video segments
-- video features
-- video shots
-- whole video object
Hierarchy Of Video Objects
• Relationship between them is both spatial and
temporal
• Spatial : represented by 2D/3D coordinate systems.
• Temporal : described by before, meet, overlaps,
during, equals, starts, finishes.
• An attempt to capture the meaning of the data itself
is described by the term semantic data modeling.
Semantic Data Modeling
• The overall approach to the department of a database
model can be described as
- Identifying useful semantic concepts
- Devise formal objects
- Devise formal integrity rules
- Devise formal operators
• In semantic analysis we attempt to describe the real
world by making simple declarative propositions
about it.
• A proposition is a sentence that describes a state of
affairs, therefore capturing something meaningful
about the system.
• 1234 – Accounts to 3010 marketing
• 2010 to Accounts
• 1234 to finance
• Problem with foreign key constraint being
violated as the primary key of department itself
is being modified.
• This is why some database developers would
prefer to model the employee and department
relationship as explicit table dep_emp.
Video object models
• Although object oriented approaches are popular,
there are specific problems with this approach
when it is applied to video data.
• This is because :
- Data ( media data and metadata – often created independently)
- OO model is too static for video data.
- Attributes associated with semantic content tend to be
discovered and added gradually to the media data.
- Descriptions of video data are user and application
dependent.
OVID : Object Video Information
Database
• Video object consists of a set of frames and their contents
can be described in a dynamic and incremental way.
• The inheritance of object attributes is based on interval
inclusion relationships.
• Interval is starting frame no, ending frame no, continuous
sequence of frames.
• An OVID object can be composed of a multiple of intervals.
• OVID video object have a definition consisting of
-- an OID
-- a set of intervals
Transforming To Database Design
• Mapping classes to tables
• Mapping subtype classes to tables
• Mapping attributes to columns
Super Server Concept
• A super server concept is about a group of servers that
coordinate among themselves to serve the request from client.
• Server assignment can be
- static or dynamic
• Dynamically allocating server reduces overloading of servers and
also waiting time of client requests.
• There are two types of systems that support super
servers.
- Distributed System
- Centralized System
• In a Distributed Environment there are media files
distributed among various servers, where each server
should gather all the up-to-date information of the load
and data residing at other servers in the distributed
network.
• In a Centralized Environment, there are two types of
severs : normal server which serves user requests
super server that takes the decision of diverting
the client request to other servers if there is any
problem in serving the client’s request at a server.
Heterogeneous Distributed Multimedia
Databases
• In a heterogeneous Distributed system, there may be
different DBMS data models included.
• E.g. there may be networked, relational, object oriented,
and object relational data models with in the system.
• This may be the result of individual data sites developing
individual databases, that were later integrated.
• Data repositories linked to both meta data and media
compilers that would be responsible for interpreting the
query and composing the result set.
• Storage manager is responsible for dynamically
managing the internal schemas.
• The system usually consists of
- Multiple Users
( UI deals with presentation and manipulation of
multimedia object and interaction with user)
- Network
(provides communication mechanism between user
and server)
- Servers
(managing multimedia db and composing general
multimedia objects for users.)
Data-links Architecture
• Data links serves as a data type and is used while
creating a table in the database.
• It refers to the link or url, which points to the
source of the data in the external storage.
• Whenever an insert operation is performed on the
created table, the column which has data link as
its type, takes the link as its data.
• When a delete operation is performed, the entire
row is deleted along with the reference link
mentioned in the column for the attribute type
Data link.
• DLFM – Data links File Manager, is responsible
to maintain the link from the database table to
the external storage where actually the file is
being stored.
• DLFF – Data links File system Filter , is used to
perform any integrity constraints which are to
be imposed on the data or for filtering the file
requests according to the availability of data.
Technical Components required to support
Content Based Retrieval
Table Space
• Most relational DBMSs assign tables to some kind of storage area
and use separate specifications to relate these storage areas to usual
operating system files termed as partition, table space, database
space or db space.
• There must be a table space with a unique name in existence before
a database can be created and objects stored in the table space.
• create tablespace users datafile   
  ?/ora01/oracle/oradata/booktst_users_01.dbf? 
Size 50m;
• For a table space called USERS assigned to a database called
BOOKTST we would have a data file called
booktst_users_01.dbf.

• We used 50m to indicate that we wanted to create the table


space 50 Megabytes in size. You can also use the K symbol for
Kilobytes, and the G symbol for Gigabytes. If you want to try
to use the P symbol for a 50 Peta byte table space.
• create table space myusers datafile
   ?/u01/app/oracle/oradata/bookt
st_users_01.dbf? 
size 50m
blocksize 32k
maxsize 100m
uniform size 10M
autoextend on;
• Autoextend on:  Creating a table space with auto extend on
alleviates the potential problem of a database outage if the table
space reaches maximum extents.  will push the point of failure to
the OS file system, which is good because you never want your
database to hang because a table space cannot grow. 
• Maxsize : This is the maximum size that the data file can grow. 
We recommend creating a table space with max size unlimited to
prevent a lock-up.
• Size: This is the size of the underlying data file for the table space.
• Blocksize : Specifying a larger block size will not result in a larger
table space. 
• Uniformsize:  This specifies the size that the data file will grow
when it needs to extend.

You might also like