8/7/2019 DRISHTI report
1/26
1. Abstract:
Pair wise similarity or comparison generally refers to any process of comparing entities in
pairs to judge which of each pair is preferred , or has a greater amount of some quantitative
property . The method of pair wise comparison is used in the scientific study
of preferences, voting systems, intelligent systems or AI systems. Extending the idea of
Learning, a measure of similarity between pairs of objects is learnt which is a fundamental
problem in machine is learning. Our goal is to learn the similarity between images in pairs.
This knowledge of similarity can be used especially for storage and retrieval of images in
databases. The same knowledge so obtained can be used to prepare clusters. Several other
practical usages exist to make use of measures learned. The practical usage of similarity
between images is for storage and retrieval in image (multimedia) databases, facerecognition, computer vision and so on.
1
http://en.wikipedia.org/wiki/Preferencehttp://en.wikipedia.org/wiki/Quantitative_propertyhttp://en.wikipedia.org/wiki/Quantitative_propertyhttp://en.wikipedia.org/wiki/Preferencehttp://en.wikipedia.org/wiki/Voting_systemshttp://en.wikipedia.org/wiki/Preferencehttp://en.wikipedia.org/wiki/Quantitative_propertyhttp://en.wikipedia.org/wiki/Quantitative_propertyhttp://en.wikipedia.org/wiki/Preferencehttp://en.wikipedia.org/wiki/Voting_systems8/7/2019 DRISHTI report
2/26
2. Introduction
Learning a measure of similarity between pairs of objects is a fundamental problem in
machine learning. Pair wise similarity generally refers to any process of comparing
entities in pairs to judge which of each pair is preferred , or has a greater amount of
some quantitative property. Machine learning is a scientific discipline that is concerned
with the design and development of algorithms that allow computers to evolve behaviours
based on empirical data , such as from sensor data or databases . A learner can take
advantage of examples (data) to capture characteristics of interest of their unknown
underlying probability distribution. Data can be seen as examples that illustrate relations
between observed variables. A major focus of machine learning research is to
automatically learn to recognize complex patterns and make intelligent decisions basedon data; the difficulty lies in the fact that the set of all possible behaviours given all
possible inputs is too large to be covered by the set of observed examples (training data).
Hence the learner must generalize from the given examples, so as to be able to produce a
useful output in new cases. Machine Learning when applied to large databases is called
Data Mining, when applied to images it has applications like face recognition,
segmentation and object-based image similarity.
Google Search Engine is a very nice idea. It provides information behind the scene byauto completion of text query and gives the most visited (highest ranked) result first.
Here we provide an application DRISHTI which can be used to define considerable
image similarity and hence can be used to search for semantically similar image(s) from a
image set.
2
http://en.wikipedia.org/wiki/Preferencehttp://en.wikipedia.org/wiki/Quantitative_propertyhttp://en.wikipedia.org/wiki/Algorithmhttp://en.wikipedia.org/wiki/Computerhttp://en.wikipedia.org/wiki/Datahttp://en.wikipedia.org/wiki/Sensorhttp://en.wikipedia.org/wiki/Databasehttp://en.wikipedia.org/wiki/Preferencehttp://en.wikipedia.org/wiki/Quantitative_propertyhttp://en.wikipedia.org/wiki/Algorithmhttp://en.wikipedia.org/wiki/Computerhttp://en.wikipedia.org/wiki/Datahttp://en.wikipedia.org/wiki/Sensorhttp://en.wikipedia.org/wiki/Database8/7/2019 DRISHTI report
3/26
2.1 Objective:
a. Study of image similarity approaches and algorithms.
b. Defining a similarity measure for images.
c. Design of an application model of above study.
d. To display the result set in as per the descending relevance to query image.
e. Eliminating of identical images from result set generated by application model.
2.2 Scope:
Every system in the world is moving towards automation and intelligence. True intelligence
is achieved only when a system starts to learn by itself, when it starts to draw conclusions
from its experience. Same is we wanted for a system which could study the similarity of
image. The application that evolved from this is not meant for the general purpose user but
scope will be in specific fields like clustering of images, face recognition.
2.3 Problem in existing system:
The study of image similarity is advanced field which is still one of the most dominating
research field. No generic approach for study of image exists. No system or approach exists
which works on all categories of images. The heterogeneity in images is from image formatsto colour models even from random pixels patterns to their colours. Some approaches exists
which works for some image sets, but do not scale to large image sets. The main reason being
the slow processing of image data, internal representation of images takes a lot of physical
memory.
This work is motivated by a research paper which aims at scaling the similarity learning in
images to large sets. This done by three important hits-
a. Similarity Learning
b. Sparse representation of images
c. PA algorithmic approach
3
8/7/2019 DRISHTI report
4/26
2.4 Platform:
2.4.1 Hardware Specification:
Table 1
2.4.2 Software Specification:
Table 2
Implementation Language: The implementation language used is MATLAB .
MATLAB is a high-level, fourth generation language and interactive environment that
enables programmer to perform computationally intensive tasks faster and easily than with
traditional programming languages such as C, C++, and Fortran. Following are the primary
reasons which led us to use MATLAB
, these are-a. Support for both Procedural and Object-Oriented paradigm support.
b. Platform Independence.
c. Specialised package for image processing
d. Database toolbox for database maintenance
3. System Requirement Analysis:
4
S.No. Description Alternatives (If available)
1. 1 Gb of RAM (for best performance) Not Applicable
2. Processor Pentium-4 or above Not Applicable
S.No. Description Alternatives (If available)
1. Windows Family
(Xp, vista, 7)
Linux family
(Ubuntu)2. MS-Access MySQL, Derdy,Oracle,Ingress
3. Matlab 7.0 Not-Applicable
4. Environment Matlab Runtime Environment
8/7/2019 DRISHTI report
5/26
3.1 Information Gathering:
Information Gathering refers to the act of understanding the nature of program to develop
an application. In our project information gathering mainly includes analyzing different
standalone software for better and friendly interface and also in order to gather information about different fields to design the system. It mainly includes
a. About Functionalities
b. About Database
c. Understanding Information Domain
a. About Functionalities: Here we analyze the functionalities which may not be
important for a general user but are crucial from learning point of view.
b. About Database: Data regarding different entities considering relationships has
been gathered. Database will be designed accordingly.
c. Understanding Information Domain: Proper understanding of information domain
will lead to a proper flow of information from database to user to & fro by appropriate
functionalities.
3.2 System Feasibility:
5
8/7/2019 DRISHTI report
6/26
3.2.1 Economical Feasibility: Economic feasibility check is an evaluation process carried
out to determine development cost and profit incurred after deployment.
In our project:
a. Software: The software used in developing this project is either provided by
the college or are provided as freeware by software websites or are cracked or are made
available for students.
b. Hardware: The configuration cost for the system is negligible as all the
hardware needed and also internet connectivity is already being provided to us by the college.
3.2.2 Technical Feasibility: Technical feasibility of this system mainly depends on thesimilarity evaluation algorithm and parameters
a. Language MATLAB
b. Database Tool MS-Access
3.2.3 Behavioral Feasibility: Our project is mostly a computational application . It is
practically not possible to achieve the accuracy at finest level but the project can be done
within given time period.
6
8/7/2019 DRISHTI report
7/26
4. System Analysis:
4.1 Information Flow Representation:
4.1.1 ER Diagram:
The entity-relationship (E-R) data model perceives the real world as consisting of basic
objects, called entities, and relationships among these objects. It was developed to facilitate
database design by allowing specification of an enterprise schema, which represents the
overall logical structure of a database. The E-R data model is one of several semantic data
models; the semantic aspect of the model lies in its representation of the meaning of the data.
The E-R model is very useful in mapping the meanings and interactions of real-world
enterprises onto a conceptual schema. Because of this usefulness, many database-design tools
draw on concepts from the E-R model.
FIGURE 1. ER-DIAGRAM
7
Admin Image
Img_id
Img_size
Img_type
Img_name
Img_date
reg
ulates
password
8/7/2019 DRISHTI report
8/26
4.1.2 Class Diagram:
A class diagram is a type of static structure diagram that describes the structure of a system
by showing the system's classes, their attributes, and the relationships between the classes.
Class diagrams are widely used to describe the types of objects in a system and their relationships. Class diagrams model class structure and contents using design elements such
as classes, packages and objects. Class diagrams describe three different perspectives when
designing a system, conceptual, specification, and implementation. These perspectives
become evident as the diagram is created and help solidify the design
Classes are composed of three things: a name, attributes, and operations. Below is an
example of a class.
FIGURE 2. CLASS DIAGRAM FORMAT
Class diagrams also display relationships such as containment, inheritance, associations and
others. The association relationship is the most common relationship in a class diagram. The
association shows the relationship between instances of classes. For example, the class Order
is associated with the class Customer. The multiplicity of the association denotes the number
of objects that can participate in then relationship. Another common relationship in class
diagrams is a generalization. A generalization is used when two classes are similar, but have
some differences.
8
8/7/2019 DRISHTI report
9/26
FIGURE 3. CLASS DIAGRAM
4.1.3 Use Case:9
8/7/2019 DRISHTI report
10/26
A use case diagram is a type of behavioral diagram defined by the Unified Modeling
Language (UML) and created from a Use-case analysis. Its purpose is to present a graphical
overview of the functionality provided by a system in terms of actors, their goals (represented
as use cases), and any dependencies between those use cases. The main purpose of a use case
diagram is to show what system functions are performed for which actors. Roles of the actors
in the system can be depicted.
Use case diagrams depict:
Use cases: A use case describes a sequence of actions that provide something of measurable
value to an actor and is drawn as a horizontal ellipse.
Actors: An actor is a person, organization, or external system that plays a role in one or more
interactions with your system. Actors are drawn as stick figures.
Associations: Associations between actors and use cases are indicated in use case diagrams
by solid lines. An association exists whenever an actor is involved with an interaction
described by a use case. Associations are modelled as lines connecting use cases and actors
to one another, with an optional arrowhead on one end of the line. The arrowhead is often
used to indicating the direction of the initial invocation of the relationship or to indicate the
primary actor within the use case. The arrowheads are typically confused with data flow and
as a result I avoid their use.
FIGURE 4. USE-CASE DIAGRAM
4.1.4 Sequence Diagram:
10
8/7/2019 DRISHTI report
11/26
4.1.4.1 Search Image:
FIGURE 5. SEQUENCE DIAGRAM for Search Image Use case
4.1.4.2 See Results
11
8/7/2019 DRISHTI report
12/26
FIGURE 6. SEQUENCE DIAGRAM for See Results Use case
4.1.4.3 Discard results:
12
8/7/2019 DRISHTI report
13/26
FIGURE 7. SEQUENCE DIAGRAM for Discard Results Use case
4.1.4.4 Update-dataset:
13
8/7/2019 DRISHTI report
14/26
FIGURE 8. SEQUENCE DIAGRAM for update dataset Use case
4.1.5 Activity Diagram
14
8/7/2019 DRISHTI report
15/26
An activity diagram is a diagram that shows activities and actions to describe workflows. In
the Unified Modeling Language an activity diagram represents the business and operational
step-by-step workflows of components in a system. An activity diagram shows the overall
flow of control.
Construction
Activity diagrams are typically used for business process modeling. They consist of:
Initial node.
Activity final node.
Activities
The starting point of the diagram is the initial node, and the activity final node is the ending.
An activity diagram can have zero or more activity final nodes. In between activities are
represented by rounded rectangles.
FIGURE 9. ACTIVITY DIAGRAM
5. Design:
15
8/7/2019 DRISHTI report
16/26
8/7/2019 DRISHTI report
17/26
Actor: Those entities (people or devices) that interact with the target system by
producing or consuming information that is necessary for requisite processing.
The actor for our system is the administrator and general user.
FIGURE 10. Architectural Context Diagram
5.1.2 Architectural behavioural diagram: Behaviour diagrams emphasize what must
happen in the system being modelled:
a. State machine diagram: standardized notation to describe many systems, from
computer programs to business processes.
b. Use case diagram: shows the functionality provided by a system in terms of actors,
their goals represented as use cases, and any dependencies among those use cases.
5.1.3 Description of Architectural design: Architectural patterns are software patterns that
describe solutions known to work efficiently, to architectural problems in software
engineering. It gives description of the elements and relation type together with a set of
17
DATABASE PASTREFERENCES
DRISHTI
RESULT SETSIMILARITYMARKINGMODULE
ADMINUSER
8/7/2019 DRISHTI report
18/26
constraints on how they may be used. An architectural pattern expresses a fundamental
structural organization schema for a software system, which consists of subsystems, their
responsibilities and interrelations. In comparison to design patterns, architectural patterns are
larger in scale.
One of the most important aspects of architectural patterns is that they embody different
quality attributes. For example, some patterns represent solutions to performance problems
and others can be used successfully in high-availability systems. In the early design phase, a
software architect makes a choice of which architectural pattern(s) best provide the system's
desired qualities. Some architectural patterns are implemented in software frameworks which
can be used to more quickly build specific programs.
The architectural design of a standalone system is often characterized, as communication
between the database server and performing computations, in the following way:
The goal is to achieve the quality of scalability. A database server exists to provide the image
set also called training data. The user originates a call to the database server y a query image,
which is then searched for similar images, synchronously or asynchronously.
The main aspects that are considered during architectural design are:
a. The primary objective of architectural design is to develop a modular program
structure and represent the control relationships between modules.
b. Secondly, architectural design melds program structure and data structure, defining
interfaces that enable data to flow throughout the program.
5.1.4 Control Hierarchy:
Control hierarchy, represents the organization of program components (modules) and implies
a hierarchy of control. The most common notation is the three linked diagram that represents
the hierarchical control for call and return architectures. Depth and width provide an
indication of the number of levels of control and overall span of control, respectively. Fan-out
is a measure of the number of modules that are directly controlled by another module. Fan-in
indicates how many modules directly control a given module.
18ADMIN USER
8/7/2019 DRISHTI report
19/26
FIGURE 11. Control Hierarchy Diagram
5.2 Modular Approach:
Modularity is generally desirable, especially in large, complicated programs. Inputs are
usually specified syntactically in the form of arguments and the outputs delivered as return
values. Scoping is another technique that helps keep procedures strongly modular. It prevents
the procedure from accessing the variables of other procedures (and vice-versa), including
previous instances of itself, without explicit authorization. Less modular procedures, often
used in small or quickly written programs, tend to interact with a large number of variables in
the execution environment, which other procedures might also modify.
Because of the ability to specify a simple interface, to be self-contained, and to be reused,
procedures are a convenient vehicle for making pieces of code written by different people or
different groups, including through programming libraries.
5.2.1 Modules Used:
19
Authenticate
Search images
Updateimage set
Show result set Discard result set
QUIT
8/7/2019 DRISHTI report
20/26
Modules are typically incorporated into the program through interfaces. The various modules
used in our application are as follows:
a. Home page: The home page of the application is also the query page, which includes
the functionalities provided to user as well as admin.
5.2.2 Internal Data Structure: We have employed primitive data types to hold the data for
loops, counters. Images which is the external data, is represented in application using
matrices. Not as the ordinary matrix but as sparse matrix.
5.2.3 Algorithm Design for Operations:
The application so designed to learn the issue so called pair wise similarity in images is a
two-tier application, which has an user-level interface and a database holding images.
algorithm:
1. User enters the query.
2. Application program analyzes the query image.
3. Comparison with other image or images metadata-templates.
4. Create the result set
5. Display the result to user.
5.3 Data Design:
The data design creates the model of data and information i.e. represented at a high level of
abstraction. The data objects defined during software requirements analysis are modelled
using ER diagrams. The data design activity translates these elements of requirement model
into data structure at component data level.
1. Admin entity:
20
8/7/2019 DRISHTI report
21/26
FIGURE 12. Admin Entity
2. Image entity:
FIGURE 13. Image entity
5.4 Interface Design:
The interface design is the bridge for interaction between a human and a computer .It creates
an effective communication between the human and the computer .The interface design
begins with the identification of the user, task and environmental requirements .It is theinformation representation of the available data thus, if the representation is confusing or
misleading users may misunderstand the meaning of the information. The best possible
representation of the data or the easiest available interface is the Graphical User Interface
(GUI). The GUI provides us with windows, icons, menus, pointers etc. The basic interface
design principles are:
a. User familiarity:
21
8/7/2019 DRISHTI report
22/26
The interface in this project is easily understood by the user that is the
interface has the user friendly terms. In our project we have used JSP ,
HTML to build various Intranet interfaces all arranged in the page so as
to make the user much comfortable.
b. Consistency:
The project if requirement specific needs to be consistent and so is this
project which satisfies the consistency criteria as a whole. It does not
break down if any unconditional problem is there.
c. Minimal surprise:
The user must not be surprised by the systems behaviour .The system
must be capable of handling the abrupt data. Even if the system shuts
down that should not be abrupt. No abrupt switching off or anomalous
behaviour is there with the system.
5.4.1 Human-machine Interface Design specification: In the industrial design field
of human-machine interaction , the user interface is (a place) where interaction between
humans and machines occurs. The goal of interaction between a human and a machine at the
user interface is effective operation and control of the machine, and feedback from the
machine which aids the operator in making operational decisions. Examples of this broad
concept of user interfaces include the interactive aspects of computer operating systems ,
hand tools , heavy machinery operator controls. and process controls. The design
considerations applicable when creating user interfaces are related to or involve such
disciplines as ergonomics and psychology .
A user interface is the system by which people ( users ) interact with a machine . The user
interface includes hardware (physical) and software (logical) components. User interfaces
exist for various systems , and provide a means of:
a. Input , allowing the users to manipulate a system, and/or
b. Output , allowing the system to indicate the effects of the users' manipulation.22
http://en.wikipedia.org/wiki/Industrial_designhttp://en.wikipedia.org/wiki/Industrial_designhttp://en.wikipedia.org/wiki/Human-computer_interactionhttp://en.wikipedia.org/wiki/Operating_systemhttp://en.wikipedia.org/wiki/Operating_systemhttp://en.wikipedia.org/wiki/Toolshttp://en.wikipedia.org/wiki/Toolshttp://en.wikipedia.org/wiki/Heavy_machineryhttp://en.wikipedia.org/wiki/Unit_operationhttp://en.wikipedia.org/wiki/Unit_operationhttp://en.wikipedia.org/wiki/Ergonomicshttp://en.wikipedia.org/wiki/Psychologyhttp://en.wikipedia.org/wiki/Psychologyhttp://en.wikipedia.org/wiki/User_(computing)http://en.wikipedia.org/wiki/Interactionhttp://en.wikipedia.org/wiki/Machinehttp://en.wikipedia.org/wiki/Systemhttp://en.wikipedia.org/wiki/Inputhttp://en.wikipedia.org/wiki/Inputhttp://en.wikipedia.org/wiki/Outputhttp://en.wikipedia.org/wiki/Outputhttp://en.wikipedia.org/wiki/Industrial_designhttp://en.wikipedia.org/wiki/Human-computer_interactionhttp://en.wikipedia.org/wiki/Operating_systemhttp://en.wikipedia.org/wiki/Toolshttp://en.wikipedia.org/wiki/Heavy_machineryhttp://en.wikipedia.org/wiki/Unit_operationhttp://en.wikipedia.org/wiki/Ergonomicshttp://en.wikipedia.org/wiki/Psychologyhttp://en.wikipedia.org/wiki/User_(computing)http://en.wikipedia.org/wiki/Interactionhttp://en.wikipedia.org/wiki/Machinehttp://en.wikipedia.org/wiki/Systemhttp://en.wikipedia.org/wiki/Inputhttp://en.wikipedia.org/wiki/Output8/7/2019 DRISHTI report
23/26
Generally, the goal of human-machine interaction engineering is to produce a user interface
which makes it easy, efficient, and enjoyable to operate a machine in the way which produces
the desired result. This generally means that the operator needs to provide minimal input to
achieve the desired output, and also that the machine minimizes undesired outputs to the
human.
5.4.2 I/O forms:
1. DRISHTI home page
FIGURE 14
2. Selection of query image
23
http://en.wikipedia.org/wiki/Human-computer_interactionhttp://en.wikipedia.org/wiki/Human-computer_interaction8/7/2019 DRISHTI report
24/26
FIGURE 15
6. Limitations:
Following are the limitations so far identified:
a. The application so designed will train the standalone system.b. Not for general user
c. The slow image processing has led to high response time
7. Future Scope: Though we have tried our best to make the project a success but still
there is scope for improvement.
a. Can be extended to a web or distributed application.
24
8/7/2019 DRISHTI report
25/26
b. Multiple systems can be trained simultaneously.
c. Can be availed to general user for better training and results.
8. Conclusion:
The application would comprise of all the functionalities which are discussed so far. They are
as follows-
a. Train the system to learn the image similarity
b. Search for most relevant images
c. Elimination of duplicity.
9. Bibliography and References: Provided that there are important resources that are
used as references while preparing the report, a complete list of the titles of references
concern must be included.
Help from my project guide Mr. Sapan Prajapati.
From Mr. Preetesh Purohit, our HOD of Computer Science Department.
http://citeseer.ist.psu.edu/ http://en.wikipedia.org/wiki/Machine_learning
http://ai.stanford.edu/~nilsson/mlbook.html
http://en.wikipedia.org/wiki/Matrix_norm
http://www.wisegeek.com/what-does-passive-aggressive-mean.htm
http://www.google.co.in/url?
sa=t&source=web&cd=1&ved=0CBYQFjAA&url=http://www.cs.technion.ac.il/~ela
d/publications/journals/2006/KSVD_Color_IEEE_TIP.pdf&ei=ypbqTOvQIYzvcJTE-fgK&usg=AFQjCNHEvACt3IRw0MRoqw_8DeiO6Ect_A
http://en.wikipedia.org/wiki/Semantic_similarity
An online algorithm for large scale image similarity learning- a research paper.
Other references in the above stated paper.
10. Appendices:
25
http://citeseer.ist.psu.edu/http://citeseer.ist.psu.edu/http://en.wikipedia.org/wiki/Machine_learninghttp://en.wikipedia.org/wiki/Machine_learninghttp://ai.stanford.edu/~nilsson/mlbook.htmlhttp://ai.stanford.edu/~nilsson/mlbook.htmlhttp://en.wikipedia.org/wiki/Matrix_normhttp://en.wikipedia.org/wiki/Matrix_normhttp://www.wisegeek.com/what-does-passive-aggressive-mean.htmhttp://www.wisegeek.com/what-does-passive-aggressive-mean.htmhttp://www.google.co.in/url?sa=t&source=web&cd=1&ved=0CBYQFjAA&url=http://www.cs.technion.ac.il/~elad/publications/journals/2006/KSVD_Color_IEEE_TIP.pdf&ei=ypbqTOvQIYzvcJTE-fgK&usg=AFQjCNHEvACt3IRw0MRoqw_8DeiO6Ect_Ahttp://www.google.co.in/url?sa=t&source=web&cd=1&ved=0CBYQFjAA&url=http://www.cs.technion.ac.il/~elad/publications/journals/2006/KSVD_Color_IEEE_TIP.pdf&ei=ypbqTOvQIYzvcJTE-fgK&usg=AFQjCNHEvACt3IRw0MRoqw_8DeiO6Ect_Ahttp://www.google.co.in/url?sa=t&source=web&cd=1&ved=0CBYQFjAA&url=http://www.cs.technion.ac.il/~elad/publications/journals/2006/KSVD_Color_IEEE_TIP.pdf&ei=ypbqTOvQIYzvcJTE-fgK&usg=AFQjCNHEvACt3IRw0MRoqw_8DeiO6Ect_Ahttp://www.google.co.in/url?sa=t&source=web&cd=1&ved=0CBYQFjAA&url=http://www.cs.technion.ac.il/~elad/publications/journals/2006/KSVD_Color_IEEE_TIP.pdf&ei=ypbqTOvQIYzvcJTE-fgK&usg=AFQjCNHEvACt3IRw0MRoqw_8DeiO6Ect_Ahttp://www.google.co.in/url?sa=t&source=web&cd=1&ved=0CBYQFjAA&url=http://www.cs.technion.ac.il/~elad/publications/journals/2006/KSVD_Color_IEEE_TIP.pdf&ei=ypbqTOvQIYzvcJTE-fgK&usg=AFQjCNHEvACt3IRw0MRoqw_8DeiO6Ect_Ahttp://en.wikipedia.org/wiki/Semantic_similarityhttp://en.wikipedia.org/wiki/Semantic_similarityhttp://citeseer.ist.psu.edu/http://en.wikipedia.org/wiki/Machine_learninghttp://ai.stanford.edu/~nilsson/mlbook.htmlhttp://en.wikipedia.org/wiki/Matrix_normhttp://www.wisegeek.com/what-does-passive-aggressive-mean.htmhttp://www.google.co.in/url?sa=t&source=web&cd=1&ved=0CBYQFjAA&url=http://www.cs.technion.ac.il/~elad/publications/journals/2006/KSVD_Color_IEEE_TIP.pdf&ei=ypbqTOvQIYzvcJTE-fgK&usg=AFQjCNHEvACt3IRw0MRoqw_8DeiO6Ect_Ahttp://www.google.co.in/url?sa=t&source=web&cd=1&ved=0CBYQFjAA&url=http://www.cs.technion.ac.il/~elad/publications/journals/2006/KSVD_Color_IEEE_TIP.pdf&ei=ypbqTOvQIYzvcJTE-fgK&usg=AFQjCNHEvACt3IRw0MRoqw_8DeiO6Ect_Ahttp://www.google.co.in/url?sa=t&source=web&cd=1&ved=0CBYQFjAA&url=http://www.cs.technion.ac.il/~elad/publications/journals/2006/KSVD_Color_IEEE_TIP.pdf&ei=ypbqTOvQIYzvcJTE-fgK&usg=AFQjCNHEvACt3IRw0MRoqw_8DeiO6Ect_Ahttp://en.wikipedia.org/wiki/Semantic_similarity8/7/2019 DRISHTI report
26/26