You are on page 1of 17

1

Multimedia Data Mining: A Survey


SARLA MORE1, AND DURGESH KUMAR MISHRA2
PRATIBHA: INTERNATIONAL JOURNAL OF SCIENCE, SPIRITUALITY,
BUSINESS AND TECHNOLOGY (IJSSBT), VOL. 1, NO.1, MARCH 2012

Zain shaukat
[6928]
MS-SE
CUSIT

Abstract

Multimedia data mining (MDM) can be defined as the process of finding


interesting patterns from media data such as audio, video, image and text .

Since it is not generally accessible by basic queries and associated results.

MDM is the mining of knowledge and high level multimedia information from
large multimedia database system

It involves

pattern discovery,
rule extraction and
knowledge acquisition from multimedia database

Abstract

Compareing MDM techniques with the other data mining


techniques involving clustering, classification, sequence
pattern mining, association rule mining and visualization.

This paper elaborates basic concepts, application at


various areas, techniques , approaches and other useful
areas which need to be work for MDM.

INTRODUCTION

According to MPEG-7 Standard there are four kinds of multimedia data

audio data (includes sounds, speech, and music)

image data

video data(include time-aligned sequences of images) and

electronic or digital ink, (sequences of time aligned 2D or 3D coordinates of


a stylus, a light pen, data glove sensors, graphical, temporal, relational and
categorical data or a similar device are stored in a multimedia database.

MDM is the exploration and analysis, by automatic or semi-automatic means,


of large quantities of data in order to discover meaningful patterns and rules

goals of MDM are to discover useful information from large disordered data
and to obtain knowledge from the information

Introduction

There are mainly six tasks for MDM:

summarization,
association,
classification,
clustering,
trend analysis and
deviation analysis.

The working of MDM system is similarity search in multimedia data,

1. Description based retrieval system,


build indices and perform object retrieval based on image description such as keywords(
caption, size and time of creation).
2. content based retrieval system,
3. support retrieval based on the image content (such as color, histogram, texture, shape, object
and wavelet transform).

introduction

The multimedia mining involves two basic steps:

Extraction of appropriate features from the data.


Selection of data mining methods to identify the desired information.

Data mining tool operate on structured data since powerful tools are required for
the unstructured or semi-structured data in multimedia database.

Multimedia mining reaches much higher complexity from huge volume of data such
as diversity of sensor, time or condition of acquisition.

Related work(background)

The problems of multimedia data (capture, storage, transmission, and presentation)


have been investigated in the middle of 1960where the multimedia standards

MPEG-4,
X3D,
MPEG-7 and
MX have continued to grow.

For multimedia distribution and database applications different algorithms need to be


used.

database can be queried, (e.g. with the SQL multimedia and application packages
called SQL/MM)

Related work(background)

The MDM covers the following areas:

Media compression and storage.


Delivering streaming media over networks with required quality of service.
Media restoration, transformation, and editing.
Media indexing, search, and retrieval.
Creating interactive multimedia systems for learning/training and creative art
production
Creating multimodal user interfaces

Related work
(Multimedia Mining)

Processing Text: (Unstructured text documents can be represented as bag-of-words,


naive Bayesian classifier used )

Processing Graphs: (Between the classic attribute-value and multi-relational representation


of training data graph structures are there. )

Processing Images: (e.g. Texture analysis, line detection, edge detection, segmentation
and region of interest processing. histograms )

Processing Audio: (Band energy, zero crossing rate, frequency centroid, Band-width and
pitch period are most frequently used features for audio processing, wavelet
transformation)

Processing Video: (automatic segmentation, indexing, content-based retrieval and


classification )

1.

Related work (MDM goals and


methods)

Dissecting a Set of objects

The most popular goal in data mining is


Dissecting(dividing ) a set of objects described by high-dimensional data into
small comprehensive
units,
classes,
substructures, or
parts
2.

Uncovering rules:

An association rule method is used


If goal of MDM is to be expressed as revealing interesting rules.

10

Related work (ARCHITECTURE OF


MULTIMEDIA
DATA
MINING
)
Architecture includes Extracting

11

data or metadata from the unstructured database ,


Store the extracted data in a structured database and apply data mining tools on the
structured database.

The main stages of data mining process are:

1.

Domain Understanding ( learning how the results of data-mining will be used )

2.

Data selection ( target a database or select a subset of fields or data records )

3.

Learning and Pre-processing (integrating, representing, coding )

4.

Discovering Patterns (heart of the entire data mining process association, classification,
clustering, regression etc)

5.

Interpretations (evaluate the quality of discovery and its value)

6.

Reporting and using discovered knowledge (generate new actions or products and
services or marketing strategies)

Issues in MDM

Content based retrieval in multimedia is a challenging problem

FEATURE FUSION:
Features extracted from multimedia data is an important issue, how
features should be integrated for mining and other applications.

12

APPROACHES TO MDM

13

The integration of storage and search techniques with standard data mining
methods is required for multimedia database mining

Promising approaches includes Construction of

multimedia data cubes,


The extraction of multiple features(patterns or derive knowledge) from
multimedia data, and
Similarity based pattern searching.

TECHNIQUES OF MDM

14

MDM Process Using Classification Rules (focus is on discovering the semantic structures )

The Hidden Markov Model

Detection of soccer goal shots using decision tree

MDM Process Using Clustering

organizing objects into groups whose members are similar in some way.

MDM Process Using Association Rules

discovering interesting relations between variables in large databases

MDM Through Statistical Modeling

a collection of an noted images is used to build models for joint distribution of probabilities

APPLICATIONS OF MDM

Digital Libraries

Traffic Video Sequences

Automated event analysis of suspicious movements

medical analysis

Media Production and Broadcasting

Customer Insight

Surveillance

Intelligent Content Service

Knowledge Management

15

Future direction

16

the MDM applications to grow especially in areas of entertainment and medicine.

Researchers in multimedia information systems, in the search of techniques for

Improving the indexing and retrieval of multimedia information, and


Looking for new methods for discovering indexing information.

17

Thank you !

You might also like