Professional Documents
Culture Documents
Number 3
May - June 2008
IETE TECHNICAL REVIEW
The Institution of Electronics and Telecommunication Engineers
PRESIDENT
S Narayana
VICE-PRESIDENTS
A K Agarwal P N Chopra Anita G Dandekar
PUBLICATIONS COMMITTEE
Chairman
M L Gupta
Co-Chairman
M C Chandra Mouly
Members
H O Agrawal S S Agrawal Smriti Dagur
M Jagadesh Kumar Surendra Pal Giridhar R Joshi
T K De T S Rathore S K Kshirsagar
Coopted
K M Paul K S Prakash Rao
Special Invitee
S C Dutta Roy P Banerjee
EDITORIAL BOARD
Chairman
Dilip Sahay
Members
H O Agrawal A K Bhatnagar R G Gupta
S S Motial Neeru Mohan Biswas H Kaushal
The IETE Technical Review invites readable articles, preferably without mathematical expressions, state-of-the-art review papers
on current and futuristic technologies in the areas of electronics, telecommunication, computer science & engineering, information
technology (IT) and related disciplines. In addition, informative and general interest articles describing innovative products &
applications, analysis of technical events, articles on technology assessment & comparison, new & emerging topics of interest to
professionals are also welcome. While all the papers submitted will go through the same detailed review process, short papers and
Practical Designs will receive special attention to enable early publication. For submission of articles, please see 3rd cover of this
issue. Detailed guidelines to authors may be seen on IETE Website : http://www.iete.org under the heading ‘Publications.’
Annual Subscription : Subscription and Advertising rates are available on request and also on website: iete.org
Address for correspondence :
Managing Editor, IETE, 2, Institutional Area, Lodi Road, New Delhi 110 003, Telephone : +91 (11) 43538841-45
Fax : +91 (11) 24649429, email : ietend@gmail.com; mgeditor@iete.org, Website : http://www.iete.org; http://
www.iete.info
FREE TO IETE CORPORATE MEMBERS : (Cost of Production: Rs.19.00)
COPYRIGHT
IETE Technical Review is published bimonthly by the Institution of Electronics and Telecommunication Engineers. All rights
of publication are reserved by the IETE.
Copyright and Reprint permission : Abstracting is permitted with credit to the source. Libraries are permitted to photocopy
for private use of readers.
It is the IETE policy to own the copyright of the scientific and technical papers it publishes on behalf of the authors and their
employers, and to facilitate the appropriate reuse of this material by others. Authors are required to sign an IETE copyright transfer
form before publication. This form is available online in IETE website:www.iete.org. IETE retains the authors’ and their employers’
right to reuse their material for their own purposes.
97
CONTENTS
Note : The Institution of Electronics and Telecommunication Engineers assumes no responsibility for the statements and
opinions expressed by individual authors.
SCAN
IETE Technical journal has made a few changes. Cover page design has
been changed to have a pleasing standard format. IETE headquarters is processing
for online submission and review of articles in IETE journal. The publication
department is putting in lots of efforts to make the IETE Technical Review of good
standard which may be accepted by readers as well as internationally.
In this issue of Technical Review, there are four articles.
The first article by A M Khan et al is regarding a computer based system
developed for a computerized control of a specific Beam Extraction System. This
was designed and tested according to the specification of the Beam Extraction
System. The authors claim that this design system has improvement over the
existing one. Beam Extraction System which is explained in this article is for
extracting electronic beam from desired orbit which is simple to operate remotely
and cost effective.
The second article is on Semantic Web Service Composition. This article
reviews some of the popular semantic web service composition method and
presents the same in tabular format. The systems based on semantic web require
performing many processes. However, in this article, authors Sandeep Kumar
and R B Mishra primarily deals regarding sementic web service composition
techniques.
Third article in this review is by Ambar Dutta et al. The paper present literature
survey of existing Corner Detection Algorithms developed during the last three
decades.
The fourth article in this Technical Review is by Chopade and Ghatol which
presents wavelet based coding algorithm, SPIHT for coding and compressing an
image mail data. The authors claim that this coding and decoding process is
comparatively fast.
Dilip Sahay
Chairman, Editorial Board
ABSTRACT
In this paper we report on a computer based system developed for the automated control of the Beam Extraction
System (BES) of Microtron (electron accelerator facility at Mangalore University). The BES helps in extracting the
electron beam from a desired orbit. The automated system is designed with two levels. At the system level, there is a
microcontroller along with some circuit components to place the mouth of the extraction channel at a particular orbit
so that an electron beam of a particular energy can be extracted from the microtron for experimental studies. The
computer kept in the control room away from the microtron machine (i.e. user level) passes the orbit number from
which the beam is to be extracted to the microcontroller. This automated system is very simple, easy to operate and
cost effective. It also helps in accurate setting of extraction channel and extraction of electron beam of desired
energy. The work described here is an attempt to improvise on the existing system and computerise the same for
accurate settings and control from a remote distance.
Fig 1 Cross sectional view of the electron beam pattern in the microtron
manual operation can be minimised by the computer. The microcontroller receives data from
automation of the above described system. the computer and executes the microcontroller
program accordingly to rotate the stepper motors.
Since the orbits of the electron beam are fixed
These stepper motors are connected to the mouth
and electrons in every orbit have a definite energy,
of the extraction mechanism by means of a long
we can place the mouth of the extraction channel in
shaft that goes inside the microtron cavity. Thus by
that orbit and extract an electron beam of required
controlling the angle of rotation of the stepper
energy. The required energy value is entered by the
motors, the shaft is moved inward or outward to
user in the remote computer. The simple c code
position the mouth of the extraction channel in an
accepts the value and converts it to the orbit number
appropriate orbit. Since the rotation has to be in the
where the specified energy can be obtained. The
horizontal X direction and angular Y direction, two
previously set orbit is the reference and is always
motors are used to move the mouth of the extraction
kept in the data file. The newly computed orbit
channel. The number of steps the stepper motor
number is compared with the reference orbit number
requires to move the extraction channel from one
and the difference is computed. This difference
orbit to another are stored in a look - up table in the
data is passed to the MCU. The MCU receives the
microcontroller. Hence, the microcontroller moves
data from the remote computer and energises the
both the X and Y stepper motors by the required
driver circuits to rotate the stepper motor and place
number of steps to place the mouth of the extraction
the mouth of the extraction mechanism at the
channel in the desired orbit.
required orbit.
Data from the computer carries the information CIRCUIT DESCRIPTION AND WORKING
as to which orbit has to be selected for the
extraction of the electron beam. This data is sent to The Fig 2 shows the block diagram of the
the microcontroller through the parallel port of the control system used to control the beam extraction
Data from
Computer
Synchronisation
Stepper
Clock
Motors
system of microtron. The microcontroller used is counter of the microcontroller as well as the up/
AT89C51 which is an Atmel make, low power, high down counters.
performance, 8-bit CMOS microcontroller with 4kB
of flash Programmable and Erasable Read Only The stepper motor is connected to the control
Memory (PEROM). It has 128x8 bit internal RAM, logic that causes its rotation, through the motor
32 programmable I/O lines and two 16-bit timers/ driver circuit. The stepper motor used has a 28
counters. The on-chip flash permits the program kg.cm torque with a 1.8 degree step. Thus a heavy
memory to be reprogrammed by an in-system driver circuit (capable of handling approx. 1.5A) is
programmer or a conventional non-volatile memory required to rotate the stepper motor. This
programmer. requirement is accomplished by using the 2N3055
transistors in the driver circuit. These power
This combination of 8-bit CPU with flash transistors can draw current up to 15A without any
memory on a monolithic chip is a powerful, highly destruction. The four transistors are switched on/
flexible and cost-effective solution to many off sequentially by the control logic to drive the four
embedded control applications [6,7]. The control coils of the stepper motor. When the transistor
logic consists of an up/down counter and a decoder.
conducts, 5V (Vcc) is applied to the motor coils and
The CMOS technology up/down counter CD4029 is
current flows through them creating magnetic fields.
used to count the clock for increasing the steps of
The magnetic field energy thus created is stored in
the stepper motor. This counter is controlled by a
the coils and rotates the motors. When the
single control pin for up/down counting. The TTL
transistor stops conducting, power to the coils is
series 74LS155 converts the binary count from the
cut-off and the stored magnetic field collapses. Due
counter to its decimal equivalent. The stepper
motor being a sequential device requires the binary to the collapsing of the magnetic field a reverse
count of the counter to be decoded to its decimal voltage (called inductive kick-back or back EMF) is
value. The output of the decoder is connected to the generated in the coils. The back EMF can be more
driver circuit. Thus, the up/down count of the than 100 Volts and can destroy the coils itself by
counter is converted to sequential switching of the burning them. Thus, diodes are required to be
driver circuits that rotate the stepper motors. An connected across the coil in reverse direction to
external synchronisation clock circuit has been absorb the reverse voltage spike. This voltage if
used to provide synchronised clock for the timer/ not absorbed by the diodes, may produce opposite
Send the
Is
complement of the
the
difference to the Yes
difference
microcontroller
negative?
No
torque and cause improper rotation of the motor is “1”, the data is for backward rotation. The required
and also damage the transistors. Any type of count is loaded into the timer /counter registers of
rectifier or switching diodes of appropriate current the microcontroller that generates the required
rating and reverse voltage breakdown rating can be delay. This delay at the microcontroller pin o/p
used. However the commonly available diodes the enables the up/down counter to count the steps of
1N4007 have been used here. rotation of the stepper motor. This count is
converted into a sequence of pulses that rotate the
The AT89C51 microcontroller is programmed stepper motor. The Figs 3 and 4 show the flow
using Atmel’s flash programmer. One step rotation diagrams of the programs for the working of the
of the stepper motor used in the circuit is 1.8o. system.
When the motor is programmed for 200 steps, it
makes one complete rotation, i.e. 360o. The look- RESULTS
up table is maintained at address location f000h to
count the steps of rotation for X and Y direction After a careful study of the requirements and
motors. The data is checked for forward / backward specification of the Beam Extraction System of the
rotation by checking the MSB of the data. If the microtron, the circuit described above was
MSB is “0” it is for forward rotation. Else if the MSB designed, tested and was found to operate
Check
Complement the offset Yes
if
value.
MSB = ‘1’ ?
No
according to the required specifications of the information and support. We also thank the technical
microtron. This circuit was first tested by simulating staff of the Microtron Centre, Mangalore University, for
it using an 8031 microcontroller kit and later the their help and support during every phase of the work.
program was burnt into AT89C51 microcontroller.
REFERENCES
Thus, an automated control system for extraction
of electron beam of desired energy in the microtron 1. Y Sheth & B J Vaidya, Control system for Microtron
has been constructed. This system is more flexible a Mangalore University, Proc. of the Intl. conference
since the user can directly enter the value of on R&D using accelerators, p 103, September 1995.
electron energy. The control can be done even from
2. A M Khan, Mohammed Mahfooz Sheikh, B
a remote distance. The designed circuit is an
Hanumaiah, Ganesh & K Siddappa, Design of a
improvement over the existing circuit where the Computer based Control System for the Cathode
distance by which the mouth of the extraction Power Supply of Microtron, Proc National Symposium
channel has to be moved is entered manually. on Instrumentation (NSI) - 30, p 165, Nov-Dec 2005.
4. A M Khan, Mohammed Mahfooz Sheikh, B 6. Muhammad A Mazidi & Janice G Mazidi, The
Hanumaiah, Ganesh & K Siddappa, Design of a 8051 microcontroller and embedded systems,
Computer based Control System for the Magnetron Fifth Indian Reprint, Pearson Education Inc.,
Power Supply of Microtron, Proc. National Delhi, 2003.
Conference on Emerging Trends in Embedded
Systems, p 1, Oct 2005. 7. Kenneth J. Aiyala, The 8051 microcontroller
architecture programming and applications, Third
5. S P Kapitza & V N Melekhin, The Microtron, 1st edition, edition, Penram Intl Publications, Mumbai,
Hardwood Academic Publishers, London, 1969. 2003.
Authors
A M Khan obtained his MSc in Applied Electronics from Gulbarga University and his PhD from
Mangalore University. He is presently Reader & Chairman, Dept of Electronics, Mangalore
University. He has teaching and research experience of around 15 years. His research interests
include embedded systems, biomedical electronics and communications.
Address: Department of Electronics, Mangalore University, Mangalagangori 574 199.
email: <asifabc@yahoo.com>
* * *
Mohammad Mahfooz Sheikh obtained his MSc in Electronics from Mangalore University. He is
presently working as a research fellow in the Dept of Electronics, Mangalore University. He has
also worked at National Aerospace Laboratories, Bangalore as a graduate trainee. He is also
pursuing his PhD studies in the area of optical switching.
email: <ganeshsanjeev@yahoo.com>
* * *
K Siddappa is former Director, Microtron Centre and former Vice-Chancellor, Bangalore
University. He is presently the Honorary Director, JSS Foundation for Science and Society,
Bangalore. He has teaching and research experience over three decades.
ABSTRACT
The research community of Web, presently is working to generate its next generation i.e. Semantic Web. They are
moving towards automation of the retrieval and processing of web contents. The systems based on the semantic web
require performing various processes like discovery, selection, composition, orchestration, monitoring etc. on
services provided by it for satisfying client needs. In this paper, we will be concentrating on semantic web service
composition techniques. The work deals with exploring different type of composition techniques, categorization of
them, and comparing them based on some of their properties like process, reasoners & languages involved, interface
etc.
groups based on their corresponding approach used respectively. This tool exploits AI planning
i.e. AI planning based, workflow based, ontology approach based on the “Planning as Model
based, agent based and other approaches. Other checking” framework [3] and MBP tool [4]. WS-
approaches consist of approaches like iterative, GEN is composed of five software modules –
logic based, template based, interactive OWL2STS, BPEL2STS, COMPOSE, MBP Planner,
composition methods, multilevel Composability. and STS2BPEL.The tool firstly describe the existing
web services specified in OWL-S or BPEL4WS into
2.1. AI Planning based state transition system using first and second
software modules. Then these described state
The AI planning based service composition transition systems for the existing web services
consists of firstly generating the plan for are used by the COMPOSE module for composition
composition before performing the actual of goal service. The MBP planner is used to
composition. generate a plan to control the interaction with
external existing services in such a way that the
Java based Automatic Composition result is according to the goal service. STS2BPEL
then translate the result described in state transition
A Java based completely automatic semantic system form into BPEL4WS executable process
web service composition approach [1], not requiring and also provide information for execution and
any human-intervention, has also been developed deployment of resulting processes.
which uses planning-based approach for handling
protocol heterogeneity problem and service-based CLM Formal Model
approach for handling data heterogeneity problem.
The loop generation in planning is based on the An AI planning-oriented formal model [5],
pattern-based approach. The system extends Casual Link Matrix (CLM), for functional
GraphPlan, an AI planning algorithm, for generating composition of the semantic web services has
the control flow of processes by considering the been used. This model uses a regression based
structure of input and output messages in addition approach for composition of web services using
to the precondition and effects of operations. The Ra4C algorithm and perform composition of web
system generates executable workflow directly in services by automating the process of chaining of
BPEL using IBM BPWS4J API and which is run on web services based on their functional description
Oracle BPM engine. The available message to be according to the casual link (a semantic link). The
input in the BPEL process is converted into the casual link gives a logical dependency among input
required format using data mediation service. and output parameters of different web services.
However, if the available message has more than The CLM pre-computes all casual links between
one candidate element, the best one is selected by web services as an Output-Input matching. Then it
using a context-based ranking algorithm. The stores all these casual links in order to find the best
ontologies in the system is handled using Jena and web service composition. Out of this, a
uses matching or mapping techniques for handling semantically well ordered and linked plan of web
the multiple ontologies involved. This technique is services forms the solution of web service
very flexible in adapting to the new scenario, as it composition.
only needs to adapt the task specification and
discovery rules & preferences while moving to the
new scenario. 2.2. Workflow Based
OWL-S Process Model parameters. The workflow starts by finding all the
services with the input matching with requested
An OWL-S process model using workflows [6] input. Then the services are added into the workflow
for composition has been proposed. The model with the inputs matching with the outputs of the last
describes the required composite services in the service in the workflow. This process of adding
terms of other constituent services. This model is services to generate a workflow-graph with each
based on the DAML-S Virtual Machine Matchmaker branch representing a workflow continue until no
Architecture. All the processes in the environment more matches among the unused services can be
are categorized into three classes as Atomic, found or the outputs of the last service in the
Simple, and composite. In this model, the workflow match with the requested outputs. It is to
composition process is defined as a hierarchy of note that the each branch holds its own list of
workflows. The workflows are constructed using unused services. The process described here is in
atomic, simple or other composite processes using forward chaining fashion. However, this approach
various composition constructs like if-then-else, also allows composition using backward chaining
repeat-until, repeat-while, iterate, split, split+join, fashion. The output here can be conditional as well
sequence, unordered, choice. The model can be as unconditional. In the case unconditional output
used reasoning of the possible compositions and is there, both the forward as well as backward
also to control the invocation of services. It also chaining approach proves equally good, but in the
uses a special application tool WSDL2DAMLS. case of conditional output, the backward chaining
approach is usually used.
Knowledge based Approach
OntoMat- Service Infrastructure
A knowledge based semantic web service
composition approach [7] implemented in a An OntoMat-Service infrastructure [9] which in
workflow construction environment has been addition to having easy to use user-interface also
proposed which uses domain knowledge for service provides capability to generate reasonably complex
composition, selection and instantiation process. workflows has been developed. It uses a particular
The system generates workflows which can be type of semantic annotations called deep
executed using a domain-specific direct mapping annotations [10]. The OntoMat-Service browser
mechanism as well as using WSDL-based service provided by it has enhanced features like direct,
grounding. For generating the workflow manual invocation of an advertised service,
specifications, in addition to the domain knowledge, invoking aggregated web services. The OntoMat-
it also uses semantically enriched service Service system has web services with the service
descriptions which assist in the process of description using WSDL (Web Service Description
discovery which further leads to the workflow Language). The web services are presented before
specifications. A framework implementing this the user in the form of a nicely formatted HTML/
approach consists of a Workflow Construction XHTML document, which can be seen of the user at
Environment, a set of web services, knowledge the client side using OntoMat-Service browser.
bases and ontologies. The framework mainly This browser also uses deep annotations to
emphasis the use of DAML-S for representing the highlight the human-understandable items
service descriptions. associated with underlying machine-
understandable semantics. The user at the client
Information transformation on I/O side can not only view the web services but also the
parameters mapping rules. User here selects a set of web
service operations and mapping rules. These
A composition approach [8] in which services is selections are then used by the web service planner
composed into workflows has been proposed. In module to compute the possible web service
this approach, the composition into workflows is workflows using pre-conditions and post-conditions
based on the information transformation the of web services. However, this model is not much
workflows undertake in their input/output efficient for building the complex workflows.
dynamic composition which can cop with changes providers. In the semantic web, queries are
in environment affected by service execution flow processed by matchmakers to find appropiate
has been proposed [25]. In this system, an agent services among these advertised using semantic
based middleware layer, as a service mediator, is web language descriptions. Abstract capability
required for scalable intelligent dynamic service descriptions in matchmaking queries is more
composition interpreted as tasks comprising of effective in trade off mechanism. Clients do
varying granularity. It is composed by dynamic negotiation and filtering, in with discovery which
coalitionof service providing agents (SPA) in the involves catched revised capability
participating task.It basically decomposes the advertisements.
incoming task on the basis of their local knowledge/
Service Engagement: Negotiation and
ontology. Based on iterative extended iterative net
contracts- both the agents (service provider and
negotiation arranged for activity outsourcing to
requester) are engaged on the agreement on some
other SPAs by providing service context relocation.
terms and coditions and compromised values. The
However there are some of main issues still
functional requirment depends on the interaction
unsolved or partially solved as like no common
capability, in general and in particular on service
mechanism for activity outsourcing activity
request formulation, contract prelimitriaries and
parameters for adjustment meaning of negotiation
contract negotiation. The negotiation affect the
ontologies family, unsufficient methodology for
architetures requirements such as : negotiation
representation of task /activities /services dynamic
protocols and and services and auditing services.
structure at coarse level and task/process
The service engagement protocols describe
ontologies with fine granularity, non standardization
messages exchange between providers and
of measure, assesment and evaluation of human
requesters requesting the agreements. The
base parameters such as trust, crediability,
equivalent protocols are implemented in FIPA. The
reliability andcapability. In the system , semantic
temporal flow and semantic understanding of
fact of request task activity service layering activity
negotiations that can occur between two agents are
context is translated with DAML-S markup
incorporated in the negotiations committment
corressponding to service profile; the service is
protocol.
than invoked via the internet specified by its binding
i.e. grounding in DAML-S description. Enactment: After engagement the entactment
is fact to be in contact and deal on agreement in
Interactive Agents space and time (temporal) continum. The functions
are response interpretation, translation
The agents in semantic web services are used interpretation, execuation, process mediation and
as clients (requesters), service providers, and delegation and dynamic service composition. The
middle agents. The client specifies what they want middle agent has to follow certain protocols and
from services. Client will need to avoid hard coded ontologies for interacting with user agent as :
knowledge of the syntax for interacting with process mediation services , process scheduling
providers. It is to mediate to classes of functionality and composition services. The interoperativity
similar providers by using information from would be enhanced by the adaptive components
published semantic service description. Three main built up by sharing of abstract models rather than
functions related to composition of semantic web syntactic and abstract agreements among
services in this method are service discovery, developers.
engagement and enactment [26].
Behavior (BDI) Agent
Service Discovery: In this process agent
(clients) identifies required sevices to achive its In the Semantic web, the web pages are marked
objectives. There are three steps in the process : up in accordance with standardized conventions in
service providers who perform the services, order to reduce ambiguity and facilitate automated
services requesters seek services that can reasoning. Web Services is not simply information
accomplished an internal objective, and but behavior which can be viewed as the
matchmaker performs matching between client fundamental attribute of intelligence. An agent
requirement and available services from it its should have the following attributes as goals,
intentions, beliefs, and behavior. Behaviors are the by using JENA and the OWL-S based ontologies
actions the agent is able to take. The process are processed using Mindswap’s OWL-S APIs. The
model in DAML-S specifies the behavior of the user interface of the system is graphical & user
service [27]. Behavior oriented design is one of the friendly and is implemented with the help of some
approaches to model agent behavior consisting of commercially available toolkits like JFACE and
action, perception and learning. The action of an SWT (Standard Widget Toolkit).
agent is supported by the perception composed of
knowledge and beliefs to determine how and when Iterative Approach
action should be expressed. The single behavior
module may be determined with actions dependent A framework, MoSCoE (Modeling Service
on state and sensing action that mention the Composition and Execution) [29], based on iterative
accuracy of the state. The BOD modules are approach for service composition has been
incorporated into hierarchical reactive plans into proposed, in which the incomplete goal service
the system execution. It includes a specification for specifications are reformulated iteratively for
parallel rooted ordered slip–stack Hierarchical composition until the desired goal service is
(POSH) reactive plans. The BOD modules contain obtained. Services in this framework are modeled
encapsulated state provide the overall agent with using Symbolic Transition System (STS). So in this
perception primitives. BOD has been implemented system, all the component services and goal
in DAML-S. services are modeled as STS. The system starts
with a choreographer and an abstract STS of goal
2.5. Miscellaneous service. Then using given component services, a
subset is identified and composed with
In this section, we will consider examples for choreagrapher to create goal service. However in
some other used approaches of semantic web the case of failure to realize goal service, it modify
service composition. We will consider cases for the some specific states and transitions of goal
context based, iterative, BPEL4WS, template service STS and thus reformulate the goal
based, and logic based approaches for service specifiction iteratively until a successful
composition. composition is obtained. This framework also
provide the user with ability to abort the process
Context Based when required.
partners and implement some features defined in Linear Logic theorem Prover and Semantic
BPEL4WS specification). Reasoner operates together for achieving the goal.
The system have a translator (transformation
Template Based between internal and external presentation of web
services), Linear Logic Theorem Prover (check
A semantic web service composition approach whether user’s request for service can be
based on the Semantic Process Templates (SPT) achieved), semantic reasoner (detecting
has been proposed which uses these templates for relationship between concepts), and an Adapter
capturing the semantic requirements of the (transformation between Linear Logic and
processes and thus allowing dynamically changing Description Logic [34]). For composing the web
the services and partners [32]. SPT uses ontologies services, the user’s request for composite service
for describing the activities requirements in the as well as description of existing web services are
templates and thus provide good method of translated into sequent and extralogical axioms of
describing and locating the services. The LL respectively. Then adaptor requests semantic
ontologies are build and processed using Jena reasoner to analyze subtype relations of classes
toolkit and the composition of the semantic web and properties and covert them into LL axiom forms,
process templates are done using a METEOR-S which are send to the LL theorem prover. The LL
Web Service composition Framework tool i.e. theorem prover generates the process calculus of
Process Builder. Generic process templates are composite service after checking whether
created based on the list of activities and control requested service can be generated or not, which
flow constructs to link the activities. A WSDL (Web further is converted to DAML-S service model or
Service Description Language) editor is used by BPEL4WS.
the framework for describing the desired processes
in WSDL. The activities are then added to the A multilevel Composability Model
process templates opened or created by the
process designer tool. Also, if needed, the control Composibility is checked through a set of rules
flow constructs may also be added to the templates. organized in four levels: Syntactic, Static semantic,
However activities may also be annotated. Now for dynamic semantic and qualitative levels [35]. The
each of the activity some of the services are web services operations (concepts) are defined by
discovered, ranked and selected based on some a meta-data ontology which provides concepts that
discovery, matching and Quality of Service (QoS) are the description of other conceptions. The
criteria. Data flow is maintained between the concepts (services) are defined by a set of
selected services and using a process generator functional and non-functional attributes. The
tool which uses WSDL description of processes, functional attributes are syntactic, static semantic
process templates and WSDL description of and dynamic semantic. The non functional or
existing participating service, the executable qualitative attributes are time availability and cost.
processes are generated. These executable The syntactic attributes represent the structure of a
processes can then be deployed and invoked after service operation such as input and output
the validation. parameters that define operation’s messages. The
meaning of operations or its messages is the
Logic Based semantic attributes which are of two kinds: static
and dynamic.
A Java based automatic service composition
architecture based on Linear Logic (LL) Theorem Horizontal and vertical composition: Horizontal
Proving and having graphical user interface has composition is like a supply chain combination of
been introduced [33]. The composite services in operations. For example, a translation of German
the system are represented using Process Calculus to Hindi can be performed by horizontal composition
which is attached to the Linear Logic inference of translation from German to English and then
rules. In the architecture, the services are internally from English to Hindi. Vertical composition refers
presented using ExtraLogical axioms and Linear to the subcontracting a services to another
Logic proofs while externally they are presented services. X has a service to perform but it gives it
using DAML-S. All the three, DAML-S translator, Y to perform and Y returns it to X after performing
the services. Composability degree refers to the • The input, output, precondition and effect
weight assigned to the syntactic and semantic [IOPE] of a service component enhance the
composability. The weight signifies the importance filtering process.
of a level or rule. A minimum threshold value of
composability (0 < τ < 1) designates the services The semantic annotations of services enable
for composability. service composition. Two web services one
language (German to English) translator and
Semantic Description of Services: Static
another dictionary (English to German) can be
semantics gives various attributes at lowest level
composed together to a user’s requirement of
of granularity like serviceability, provider and
German Dictionary. Similarly, another example for
consumer types, category and purpose of different
service composition is of sensor data acquisition,
operations. Static semantics for messages consist
sensor data processing.
of the set of attributes to model the semantics of
message parameters like data type, business role,
Semantic Service Description: It is described
unit and language. Dynamic semantics or business
in OWL-S language with three components. Service
logic of an operation refers to the output expected
profile which describes the services to be
after execution of an operation, at the given specific
performed by specifying input and output types,
condition. The business logic is defined by a set of
preconditions and effects (IOPE). It is analogous to
rules where each rule has the format
yellow page like advertisements in UDDI. The
(Pre Condition Slkm, Pre Condition ikm) process model describes the working of a service
Rm = ⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯ either as an atomic process that executes directly
(Post Parameters Slkm, Post Condition ikm)
or composite process that is composition of sub
Where Pre Parameters ikm and Post Parameters processes. It is similar to business process model
ikm are sets of parameters, each parameter is in BPEL4WS.
defined by name, data type, business rule, unit and • The grounding refers to the access and
language of different services (a kind of service communication protocol of an agent;
composition). It is the discretion of the user to parameters to use in the protocol, and the
combine the services from the sensor (at different serialization technique in the communication.
physical location), or perform the pre-processing It is mapping from OWL-S to WSDL.
steps and deploy the pattern recognition services
(statistical or syntactic). • All of the OWL’s domain modeling features
can be used to directly structure the OWL-S
Interactive Composition Techniques service descriptions as well as the concepts
A goal oriented, interactive composition from other ontology.
approach, driven by filtering and selection of
services has been developed [36]. This is based on The OWL-S ontology may consist of top level
incremental generation of composition by forward class sensor and subclasses as Acoustic sensor,
and backward chaining of services. At each infrared sensor, and microwave sensor. Service
incremented step, new services are added and profiles are associated with the root class i.e.
matched to current services requirement, enabling sensor profile and subclasses with an Acoustic
the filtering of undesired services based on the Sensor profile etc. The extensible service
user’s decision and context of the required service parameter can be used to define non IOPE
profile. attributes by more specific named classes of
“Service profiles” such as “Nearby acoustic
• After filtering the compatibility of the services,
sensors”. The user can also define complex class
these are delivered as the next step of
expressions to define their requirements. The
composition process.
“Grounding” contains the information for executing
• Propinquity is determined to assess the services like pointers into WSDL description o
desirable composition. directly invoke.
3. TABULAR COMPARISON
Automatic • Convert the available message into Jena Java, BPEL Graphical Oracle
Service required format for BPEL process. User BPM
composition • Select the best one candidate element, interface Engine
[1] if multiple elements are there in available
message.
• Generate executable control flow of
processes by considering the structure of
input & output messages and precondition
& effects of operations.
WS-GEN [2] • Describe web services in State Transition No Specific OWL-S, Graphical MBP tool
system form. Reasoner BPEL4WS user [4]
• Generate plan to control the interaction with interface
existing services.
• Compose services according to goal service.
• Translate the resulting state transition
system goal service into executable
process.
Functionality • Computes casual links between web No Specific OWL-S No specific Ra4C
based services using CLM. Reasoner inteface Algorithm
semantic • Store all casual links. tool
web service • Create semantically ordered and linked
composition plan of web services.
[5]
Workflow based
Knowledge • It works as a composition advice based on Semantic DAML-S WCI GUI Software
Based two models: Service description and based search provides facilitates agent to
Systems [7] conceptual links between services and engine is basic work flow monitor to
their properties. The KBS approach realized schema construction. service
involves identification of knowledge through for
Contd.....
intensive tasks. The domain level elicitation reasoner describing composi- process
or automatic acquisition of KW using KW FaCT or a web tion and collect
acquisition method. match service. snapshots
• The representation or modeling of domain maker of states
KW in formal, symbolic structure computing when
form. advice is
• Use of KW format user’s requirement to requested
update and maintain formalized KW.
• KBS framework consists of set of components:
Work flow construction environment, a set of
diverse web services, knowledge bases and
ontologies.
Workflow • Find all services with input matching with No Specific No specific Graphical No specific
based requested input. Reasoner Language User tool
service • Add services in Workflow with inputs matching Interface
composition with outputs of last service in workflow.
[8] • Continue until output matches with the
requested output or no unused service.
OntoMat- • Present the web services having WSDL- No Specific WSDL, Graphical No specific
Service represented descriptions before the user Reasoner HTML, user- tool
System [9] using deep annotations. XHTML interface
• Selection of service operations and mapping
rules are done by user.
• Use the selections to compute web service
workflows.
Ontology Based
Contd.....
Miscellaneous
ConWe • Consolidate the context. JENA OWL-S, Graphical JFACE,
Scproto • Conciliate contexts. OWL-C user SWL, and
type [28] – • Relate services to contexts. interface indswap’s
Context • Secure contexts. using OWL-S
based • Consolidate security contexts. commercial APIs.
• Conciliate security contexts. toolkits
MoSCoE • Identify a subset of given component No Specific No specific Graphical STS
[29]- services. Reasoner Language User (Symbolic
Iterative • Compose subset with a choreographer to Interface Transition
approach get goal service. System).
• Reformulate goal specification iteratively
until successful composition.
Adapting • Collect DAML-S or OWL-S service profiles No Specific DAML-S, No specific BPWS4J
BPEL4WS into a repository. Reasoner OWL-S inteface Engine
[15] – • Define properties of partner services.
Using • Query for the partners from services in
BPEL4WS repository.
• Integrate semantic service descriptions by
querying into BPWS4J engine.
SPT based • Create Generic Process templates based Jena WSDL No specific Meteor-S
composition on list of activities. inteface tool,
[32]- • Describe desired processes in WSDL. WSDL
Template • Add activities and control flow constructs Editor.
based to the templates.
• Discover, rank and select the services for
the activities based on some criteria.
• Maintain data flow between the selected
services.
• Generate executable processes.
Logic based • Translate the user’s request composite Using a Java, Graphical DAML-S
service services into sequent of LL. specified DAML-S, user translator,
composition • Translate the description of existing web semantic BPEL4WS interface LL theorem
[33] services into extralogical axioms of LL. reasoner prover.
• Analyze subtype relation of classes and module
properties and convert them into LL axioms. in the
• Send the LL axioms such generated to LL system.
theorem prover.
• Generate process calculus of composite
service.
· • Convert it to DAML-S model or BPEL4WS.
Contd.....
Multi Level • The different levels perform various No Specific WSDL GUI and No specific
composibilty functions on the basis of rules well Reasoner Web DG tool
[35] organized in each level. manager
• Horizontal and vertical composition. using
• Weightage to different types of composability HTML
levels by level weight vector and rule weight servlet
matrix.
• Operations and messages are static and for
dynamic semantics.
Interactive • Filtering of undesired services based on No Specific OWL-S, Graphical No specific
composition user’s decision. Reasoner WSDL User tool
[36] • Determination of propinquity to assess Interface
composition.
• Use IOPE to enhance filtering process.
• Service composition using semantic
annotations.
Contd.....
5. CONCLUSION
ICAI-2001 Workshop on Planning under Uncertainty 16. H Lauren, D Roman & U Keller, Web Services
and Incomplete Information, Seattle, USA, August Modeling Ontology – Standard (WSMO-Standard),
2001. http://wsmo.org/2004/d2/v0.2/, 2004.
5. Freddy Lecue & Alain Leger, A formal model for 17. D Sell, F Hakimpour, J Domingue, E Motta & R
semantic Web service composition, 5th International Pacheco, Interactive Composition of WSMO-based
Semantic Web Conference, Athens, Georgia, 2005. Semantic Web Services in IRS-III, in: Proc of the
AKT workshop on Semantic Web Services (AKT-
6. Liliana Cabral, John Domingue, Enrico Motta, Terry
SWS04), 2004.
Payne, Farshad Hakimpour, Approaches to Semantic
Web Services: An Overview and Comparisons, 18. Y Shoham, Agent-oriented programming, Artificial
Lecture Notes in Computer Science, Springer Berlin Intelligence, Elsevier Science Publishers Ltd, vol 60,
/ Heidelberg, Volume 3053/2004, 2004, pp 225-239. Issue 1, 1993, pp 51-92.
7. L Chen, N R Shadbolt, C Goble, F Tao, S J Cox, C 19. S D Ramchurn, D Huynh & N R Jennings, Trust in
Puleston & P R Smart, Towards a Knowledge-based multiagent systems, The Knowledge Engineering
Approach to Semantic Service Composition, LNCS, Review, vol 19, no 1, pp 1-25.
Springer Berlin/ Heidelberg, vol 2870/2003, 2003, pp
319-334. 20. Peep Kungas & Mihhail Matskin, Semantic Web
Service Composition through a P2P-Based Multi-
8. Tor Arne Kvaloy, Erik Rongen, Alfredo Tirado-Ramos Agent Environment, Lecture Notes in Computer
& Peter Sloot, Automatic Composition and Selection Science, UCM, vol 4118, 2006.
of Semantic Web Services, LNCS, Springer Berlin/
Heidelberg, vol 3470/2005, 2005, pp 184-192. 21. I Stoica, R Morris, D Karger M F Kaashoek & H
Balakrishnan. Chord, A scalable peer-to-peer lookup
9. Sudhir Agarwal, Siegfried Handschuh & Steffen
service for internet applications, In Proceedings of
Staab, Annotation, Composition and Invocation of
ACM SIGCOMM 2001, San Diego, California, USA,
Semantic Web Services, www.uni-koblenz.de/
2001, ACM Press, 2001, pp 149-160.
~staab/Research/Publications/2004/web-service-
annotation.pdf, 2004. 22. Charlie Abela, Semantic Web Services Composition,
Computer Science Annual Research Workshop,
10. S Handschuh, S Staab & R. Volz. On deep
2003, Malta Council for Science and Technology,
annotation, In Proceedings of the 12th International
Villa Bighi, Kalkara, 2003.
World Wide Web Conference, WWW 2003,
Budapest, Hungary, 2003, ACM Press, 2003. 23. JADE, http://sharon.cselt.it/projects/jade/.
11. Asuncion Gomez-Perez & Rafael Gonzalez-Cabero, 24. JTP, http://www.ksl.stanford.edu/software/JTP/.
Manuel Lama, Framework for design and composition
of SWS based on Stack of Ontologies, American 25. V Ermolayev, N Keberle, O Kononenko, S Plaksin &
Association for Artificial Intelligence (www.aaai.org), V Terziyan, Towards a framework for agent-enabled
2004, pp 1-8. semantic web service composition, International Jl
of web service research, vol X, no X, 2004, pp 1-12.
12. O Corcho, M Fernández-López, A Gómez-Pérez &
M Lama, An Environment for Development of 26. Mark Burstein, Christoph Bussler, Michal Zaremba,
Semantic Web Services, In Proceedings of the IJCAI- Tim Finin, Michael Huhns, Massimo Paolucci, Amit
2003Workshop on Ontologies and Distributed Sheth & Stuart Williams, A Semantic Web Services
Systems, Acapulco, México. http://CEUR-ORG.com/ Architecture, IEEE Computer Society , IEEE Internet
vol-71/, 2003, pp 13-20. Computing, 2005.
13. O Corcho, M Fernández-López, A Gómez-Pérez, 27. J J Bryson, D Martin, S A. McIlraith & L A Stein,
and M Lama, ODE-SWS: A Semantic Web Service Semantic web services as Behavior Oriented Agents,
Development Environment, In Proceedings of the www.cs.bath.ac.uk/~jjb/ftp/ieee-daml.pdf, 2001.
VLDB- 2003 Workshop on Semantic Web and
Databases, Berlin, Germany, 2003, pp 203-216. 28. 1 S Sattanathan1, N C Narendra2 & Z Maamar,
ConWeScprototype - Context-based Semantic Web
14. Domingue, J L Cabral, F Hakimpour, D Sell & E Services Composition, ICSOC, 05, 2005.
Motta, IRS-III: A Plateform and Infrastructure for
Creating WSMO-based Semantic Web Services, In: 29. Vasant Honavar, Algorithms & Software for
Proc. of the Workshop on WSMO Implementations Interactive Discovery and Composition of Semantic
(WIW 2004), 2004. Web Services, http://www.cs.iastate.edu/~honavar/
ailab/projects/services.html.
15. Yasmine Charif & Nicolas Sabouret, An Overview of
Semantic Web Services Composition Approaches, 30. IBM, Microsoft, SAP, Siebel Systems, Business
Electronic Notes in Theoretical Computer Science, Process Execution Language for Web Services
Elsevier Science, 2005, vol 85, no 6, pp 1-8. Version 1.1, Technical report, 2003.
31. D Mandell & S McIlraith, Adapting BPEL4WS for the 34. B N Grosof, I Horrocks, R Volz, & S Decker,
Semantic Web: The Bottom-Up Approach to Web Description logic programs: Combining logic
Service Interoperation, in: Proc. of the 2 nd programs with description logic, In The 12th
International Semantic Web Conference (ISWC2003), International Conference on the World Wide Web
Sanibel Island, Florida, 2003. (WWW-2003, Budapest, Hungary, 2003.
32. Kaarthik Sivashanmugam, John A. Miller, Amit P.
35. Brahim Medjahed & Athman Bouguettaya, A
Sheth, Kunal Verma, Framework for Semantic Web
Multilevel Composability Model for Semantic Web
Process Composition, International Journal of
Services, IEEE Transactions on Knowledge and Data
Electronic Commerce, Winter 2004-5, vol 9, no 2, pp.
Engineering, vol 17, issue 7, 2005, pp 954-968.
71-106.
33. Jinghai Rao, Peep Kungas, Mihhail Matskin, Logic - 36. E Sirin, B Parsia & J Hendler, Filtering and select-
based Web Services Composition: from Service ing semantic Web services with interactive
Description to Process Model, Proceedings of the composition techniques, IEEE Intelligent Systems,
IEEE International Conference on Web Services vol 19, bo 4, 2004, pp 42-49.
(ICWS’04), 0-7695-2167-3/04, IEEE, 2004.
Authors
Sandeep Kumar obtained BE in information technology from Haryana University with Gold medal. Presently he is a
research scholar in Computer Engineering Department, IT, BHU, Varanasi. His research interest is in the area of
semantic web.
Address: Department of Computer Engineering, Institute of Technology, Banaras Hindu University, (IT-BHU),
Varanasi 221 005, India.
email:sandeep.garg.cse06@itbhu.ac.in
* * *
ABSTRACT
Corner detection is an important step in many computer vision applications. A large number of corner detection
algorithms were already developed. An exhaustive review of the existing corner detection algorithms is, therefore,
invaluable for the researchers working in this area. In the present literature, we have found a few reviews on this area.
Most reviewers classified corner detectors into two categories – boundary-based and intensity-based. Some reviewers
classified the algorithms into two groups – template-based and geometry-based. Though we follow the first
classification, still we feel that there is a requirement of further subdivision. So, we subdivide each category into
further two subcategories – spatial-domain and transform-domain. In this paper, we have reviewed a total of 114
corner detection algorithms developed from 1977 to 2006, classified them in each of the four categories and found
that most of the works have been done in spatial domain only as compared to transform domain approaches.
algorithms, a few performance measures had Splines to develop a technique for corner detection
already been proposed in the literature [6, 7, 108]. and curve representation. Beus and Tiu [10]
We have also proposed three performance developed an algorithm based on chain-coded plane
measures – Detection Gradient, False Positive curves that eliminates the detection of spurious
Ratio and False Negative Ratio, comparable with corners using a maximum cut-off value to determine
the existing measures, which are defined below. the length of the forward and backward arms of
each point. Davies [11] proposed corner detector
| NA – ND | + | NM + NF | based on generalized Hough transform. Ogowa [12]
DG = ⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯
NA computed a symmetry measure at every point on a
digital curve and then extracted corners at the local
NF
FPR = ⎯⎯⎯ maxima of the measure. Rangarajan, Shah and
NA Brackle [13] proposed an optimal gray level corner
NM detector based on Canny’s optimal one-dimensional
FNR = ⎯⎯⎯ edge detector [14]. Rattarangsi and Chin [15]
NA presented scale-based corner detection algorithm
on planar curves. Mehrotra, Nichani and
where NA, ND, NM and NF denote the total number Ranganathan [16] described two methods for corner
of corners present in the image, the number of true detection, one was based on the first directional
corners detected, the number of missed (false derivative of Gaussian and the other was based on
negative) corners and number of extraneous (false the second directional derivative of Gaussian. The
positive) corners detected in the image, detector also computed corner angle and
respectively. In the best case, the value of DG is orientation. Bell and Pau [17] developed a corner
zero, which signifies that all the corners are detector based on Freeman’s chain-code. Liu and
correctly detected without any missed or false Tsai [18] proposed a method based on the principle
corners; the same is true for FPR and FNR. of preserving gray and mass moments. Cooper,
Venkatesh and Kitchen [19] used two different
3. BOUNDARY BASED SPATIAL approaches to detect corners in an image – one, by
DOMAIN METHODS using dissimilarity along the contour direction to
detect curves in the image contour, and the other
Freeman and Davis [8] and Rutkowski and by estimating image curvature along contour
Rosenfeld [1] detected corners by using chain- direction. Orange and Greon [20] proposed
code values. Medioni and Yasumoto [9] used B- segmentation model for the detection of corners
based on a geometrical model. Xie, Sudhakar and tangency and inflection. Zheng and Zhao [39]
Zhuang [21] presented a cost minimization implemented a parallel algorithm for detecting
approach to corner detection in which they dominant points on multiple digital curves. Shilat,
associated a corner with different cost factors Werman and Gdalyahu [40] presented a method for
capturing desirable characteristics of a corner. Pei the detection of corners along ridges/troughs and
and Horng [22] extracted corners from curvature local minima points. Mokhtarian and Soumela [41]
local maxima of the shape resulting from the nest presented a corner detection algorithm based
moving average of the original image. Seeger and curvature scale space representation in which first
Seeger [23] detected corners from a gray-level the edges are extracted from the original image
image by a real-time parameter-free algorithm. using Canny edge detector [14], and then corners
Sugimoto and Tomita [24] proposed an algorithm are found from the edge image where there is a
for the detection of various feature points (corner, local maxima of absolute curvature. Sohn, Kim and
inflection and transition). Zhang, Haralick and Alexander [42] presented a mean field annealing
Ramesh [25] described a maximum a posteriori approach to boundary smoothing for curvature
(MAP) probability technique for corner detection. estimation to improve the capability of detecting
Pikaz and Dinstein [26] proposed an approach corners. Li and Chen [43] described a corner
based on a simple decomposition of the curve into detection algorithm on planar curves as a fuzzy
the minimal number of concave and convex classification problem containing three stages –
sections to detect feature points and to smooth evaluation, classification and location. Luo, Cross
noisy digital curves. Sohn, Alexander, Kim and and Hancock [44] described a corner detection
Snyder [27] presented a constraint regularization algorithm based on the topographic analysis of a
approach to detect corners that overcome the vector potential image representation. Tsai, Hou
problem of determining the unique smoothing factor. and Su [45] presented a quantitative measure of
Koplowitz and Plante [28] proposed a scheme for corners based on the smaller eigen value of the
chain-coded curves by measuring the number of covariance matrix of boundary points over a small
links to either side of a point that can produce the region of support. Guest and Fairhurst [46]
largest digital straight line. Ray and Ray [29] described a clustering approach to corner point
presented a corner detection algorithm using an analysis in hand-drawn images. Ludtke, Luo,
iterative Gaussian convolution with constant Hancock and Wilson [47] proposed a corner
window size. Dias, Kassim and Srinivasan [30] detection algorithm using mixture model of edge
presented an artificial neural network based corner orientation. Oh and Chien [48] described a corner
detection algorithm in a 2-D image. Wang, Wu, detection algorithm by combining the generalized
Huang and Wang [31] proposed a modified contour symmetry transform (GST) operator with the
tracking and efficient corner detection algorithm parametric corner equation. Shen and Wang [49]
using bending value. Lai, Paul and Wu [32] proposed a fast corner detector using modified
presented an edge-corner detection algorithm, Hough transform. Ray and Pandyan [50] presented
called Cellular Vectorization Method. Sheu and Hu an adaptive corner detector for planar curves. Marji
[33] proposed a rotationally invariant two-phase and Siy [51] proposed an algorithm for detecting
scheme for corner detection where the candidate dominant points and polygonal approximation of
corners are detected first and then these corners digitized closed curves. Wu [52] proposed an
are reinvestigated for the global trend. Arrebola, efficient method for dominant point detection using
Camacho, Bandera and Sandoval developed adaptive bending value. Urdiales, Trazegnies,
techniques which are based on local [34] and Bandera and Sandoval [53] presented a fast
circular [35] histograms of the contour chain code. algorithm for corner detection defined at different
Ji and Haralick [36] applied statistical techniques scales by estimating the curvature of a contour in a
to detect corners from chain-encoded digital arcs. local adaptive way. Banerjee, Kundu and Mitra [54]
Ray and Ray [37] used a discrete scale-space presented a support vector machine based
kernel to detect corners on a digital arc. Tsai [38] algorithm for corner detection. Guru, Dinesh and
proposed a boundary based corner detection Nagabhushan [55] presented a fast and efficient
algorithm by developing two artificial neural corner detection algorithm by introducing a new
networks, one for detecting corners with high measure “cornerity index” for the quantification of
curvature, and the other for detecting points of the prominence of a corner point. Arrebola and
detect gray-level corners in real-time algorithm Deschenes and Ziou [101] presented an approach
using fuzzy logic. Diaz, Domingo and Ayala [88] to detect the line junctions in gray images. Alvarez,
detected two-dimensional feature point from a Cuenca and Mazorra [102] proposed a
grayscale image using the application of statistical morphological corner detection algorithm to
techniques. Smith and Brady [89] proposed SUSAN estimate corners with sub-pixel accuracy and
corner detector using a concept that each image tested the detector’s accuracy in the problem of
point is associated with it a local area of similar multiple camera calibration. Wurtz and Laurens
brightness. If the brightness of each pixel within a [103] presented a corner detection algorithm in
mask is compared with the brightness of that color images through a multi-scale combination of
mask’s nucleus, then an area of the mask can be end-stopped cortical cells. Telle and Aldon [104]
defined which has the same (or similar) brightness proposed a interest point detector for color images
as the nucleus. This area of the mask is known as based on non-linear filtering of the image. Gao, Yu,
the “USAN”, an acronym for “Univalue Segment Sattar and Venkateswarlu [105] proposed an
Assimilating Nucleus”. The value of USAN gets improved Plessey corner detector, worked in scale-
smaller on both sides of an edge and becomes even space domain, for gray level images using multi-
smaller on each side of a corner. The local minima scale analysis. Bae, Kweon and Yoo [106] used two
of the USAN map represent corners in the image. oriented cross operators, COP (crosses as ordered
Kautsky, Zitova, Flusser and Peters [90, 91] pair) to extract low-level image features, viz.,
proposed a two-stage method for the detecting the corners. Elias and Laganiere [107] proposed an
feature points in which all possible candidates are approach based on a data structure similar to
found first and then the desirable number of pyramid, but of circular levels, to determine the
resulting significant points is selected among them location and orientation of corners. Golightly and
by maximizing the weight function. Laganiere [92] Jones [108] presented an algorithm for corner
proposed a morphological corner detector using an detection and matching for visual tracking of power
asymmetric closing operator. Lin, Chu and Hsueh line inspection and proposed two performance
[93] also used mathematical morphological measures – detection rate and error rate – for the
operators for corner detection with simple integer corner detection algorithms. Mikolajczyk and
computation. Stammberger, Michaelis, Reiser and Schmid [109] presented two techniques for corner
Englmeier [94] proposed a hierarchical approach to detection invariant to scale and affine
corner detection in which the whole image is transformations. Alkaabi and Deravi [110]
convolved with only one kernel. Trajkovic and presented a fast corner detection algorithm based
Hedley [95] proposed a fast corner detection on pruning candidate corners. Zhou, Liut and Cai
algorithm using a multigrid approach to reduce the [111] improved SUSAN corner detector [89] by
computational complexity and to improve the presenting a robust and efficient corner detection
quality of the detected corners. Chabat, Yang and algorithm which is capable of detecting the features
Hansell [96] used an operator to detect the true in different contrast images automatically through
location and orientation of corners. Lv and Zhou self-adjust multi-threshold. Kenney, Zuliani and
[97] modified the cornerness measure of Kitchen- Manjunath [112] presented an axiomatic approach
Rosenfeld detector [73] based on the perception of to corner detection. Vincent and Laganiere [113]
human vision system to make the detector effective proposed a new feature point detector which first
in variable illumination scenes. Zheng, Wang and performs a simple segmentation based on the
Toeh [2] presented a gradient direction corner intensity values found in the vicinity of each
detector, derived from the Plessey corner detector, considered point, and then, it tries to fit a simple
which is based on the measure of the gradient wedge corner model to the resulting segmented
module of image gradient direction. Sojka [98] area. Cooke and Whatmough [114] proposed two
presented a reliable, robust corner detection ways – one by using genetic algorithm and another
algorithm in digital images. Ruzon and Tomasi [99] by using supervised classification techniques –
used both a region model based on the distributions towards the application of learning algorithms in
of pixel colors and an edge model for the detection corner detection. Pei and Ding [115] proposed a
of corners in textured color images. Basak and corner detection algorithm based on determining
Mahata [100] proposed a connectionist model to orientations of the adaptive vertical and tangent
detect corners in binary and gray images. axes and observing the variations of brightness
along the positive and negative directions of these found that most of the work has been carried out on
axes. Rosten and Drummond [116] used machine boundary (binary) and gray-scale images. So, there
learning for fast and high quality corner detection. is huge scope of work in the field of corner detection
that will directly operate on the color images as
6. INTENSITY BASED TRANSFORM well. In addition, any boundary-based corner
DOMAIN METHODS detection algorithm consists of finding boundary
from an image, followed by following and closing
Quddus and Falmy [117] proposed a corner the boundary, which always requires some manual
detection algorithm based on scale interaction intervention resulting in high computational time.
model using Gabor filter suitable for both binary and
gray level images with variable background. 7. REFERENCES
Pedersini, Pozzoli, Sarti and Tubaro [118] proposed
1. W S Rutkowski & A Rosenfeld, A comparison of
a wavelet-based multi-resolution corner detection corner detection techniques for chain-coded curves,
algorithm based on the analysis of multi-scale Technique Report TR-623, Computer Science Center,
Laplacian’s profile. Quddus and Gabbouj [119] University of Maryland, 1978.
proposed a wavelet-based corner detection
2. A Heyden & K Rohr, Evaluation of Corner Extraction
algorithm using singular value decomposition Schemes Using Invariance Methods, in: Proc
(SVD). Gao, Sattar and Venkateswarlu [120] International Conference on Pattern Recognition, vol
presented a multi-scale corner detection algorithm 1, Vienna, Austria, pp 895-899, 1996.
on gray images using Gabor wavelets.
3. Z Zheng, H Wang & E K Toeh, Analysis of Gray
Level Corner Detectors, Pattern Recognition Letters,
7. CONCLUSIONS vol 20, pp 149-162, 1999.
4. C Schmid, R Mohr & C Bauckhage, Evaluation pf
In this paper, we provide a literature survey on Interest Point Detectors, International Journal of
existing corner detection algorithms developed in Computer Vision, vol 37, no 2, pp 151-172, 2000.
the last three decades starting from 1977 which
5. P I Rockett, Performance Assessment of Feature
include a total of 114 algorithms. We then classify Detection Algorithms: A Methodology and Case
these algorithms into two main categories – Study on Corner Detectors, IEEE Transaction on
boundary-based and intensity-based methods, Image Processing, vol 12, no 12, pp 1668-1676,
which, in turn, are again subdivided into spatial December, 2003.
domain and transform domain methods. After this 6. F Mohanna & F Mokhtarian, “Performance Evaluation
classification, we have observed that out of these of Corner Detection Algorithms under Similarity and
114 algorithms, 54 algorithms belong to boundary- Affine Transforms,” in Proc British Machine Vision
based spatial domain, 10 belong to boundary-based Conference, Manchester, UK, pp 353–362,
transform domain, 46 belong to intensity-based September 2001.
spatial domain and only 4 to intensity-based 7. F Mokhtarian & F Mohanna, Performance evaluation
transform domain. But since transform domain of corner detectors using consistency and accuracy
methods are intensely used in various applications measures, Computer Vision and Image
in image analysis and processing and Understanding, vol 102, pp 81- 94, 2006.
comparatively much less work has been carried out 8. H Freeman & L S Davis, A Corner Finding Algorithm
in transform domain, there is a tremendous scope for Chain-Code Curves, IEEE Transaction on
of work in the field of corner detection in transform Computers, vol 26, pp 297-303, 1977.
domain. Moreover, all transform domain methods 9. G Medioni & Y Yasumoto, Corner Detection and
discussed in this survey were based on wavelet Curve Representation using Cubic B-Splines,
transform and since there are many other Computer Vision Graphics & Image Processing, vol
transforms available, e.g. Fourier, Hartley, Walsh- 39, no 3, pp 267-278, 1987.
Hadamard, Haar, Slant, Karhunen-Loeve etc, so 10. H L Beus & S S H Tiu, An Improved Corner Detection
efficiency of these transforms may be investigated Algorithm Based on Chain-Coded Plane Curves,
and noted. Though the majority of work in this field Pattern Recognition, vol 20, no 3, pp 291-296, 1987.
has been carried out in the spatial domain, still 11. E R Davies, Application of the Generalized Hough
there is scope more work in this area. We have also Transform to Corner Detection, in: IEE Proceedings
on Computers and Digital Techniques, vol 135, Part of the 12th IAPR International Conference on Pattern
E, no 1, pp 49-54, January 1988. Recognition, Jerusalem, Israel, vol 1, pp 549-552,
October 1994.
12. H Ogawa, Corner Detection on Digital Curves Based
on Local Symmetry of the Shape, Pattern 26. A Pikaz & I Dinstein, Using Simple Decomposition
Recognition, vol 22, no 4, pp 351-357, 1989. for Smoothing and Feature Point Detection of Noisy
Digital Curves, IEEE Transaction on Pattern Analysis
13. K Rangarajan, M Shah & D V Brackle, Optimal Corner
and Machine Intelligence, vol 16, no 8, pp 808-813,
Detector, Computer Vision Graphics & Image
August 1994.
Processing, vol 48, pp 230-245, 1989.
27. K Sohn, W E Alexander, J H Kim & W E Snyder, A
14. J F Canny, A Computational Approach to Edge
Constrained Regularization Approach to Robust
Detection, IEEE Transaction on Pattern Analysis and
Corner Detection, IEEE Transaction On Systems,
Machine Intelligence, vol 8, no 6, pp 679-698, 1986.
Man, and Cybernetics, vol 24, no 5, pp 820-828, May
15. A Rattarangsi & R T Chin, Scale-based Detection of 1994.
Corners of Planar Curves, IEEE Transaction on
28. J Koplowitz & S Plante, Corner Detection for Chain
Pattern Analysis and Machine Intelligence, vol 14, pp
Coded Curves, Pattern Recognition, vol 28, no 6, pp
430-449, 1992.
843-852, 1995.
16. R Mehrotra, S Nichani & N Ranganathan, Corner
29. B K Ray & K S Ray, Corner Detection using Iterative
Detection, Pattern Recognition, vol 23, no 11, pp
Gaussian Smoothing with Constant Window Size,
1223-1233, 1990.
Pattern Recognition, vol 28, no 11, pp 1765-1781,
17. B Bell & L F Pau, Contour Tracking and Corner 1995.
Detection in a Logic Programming Environment, IEEE
30. P G T Dias, A A Kassim & V Srinivasan, A Neural
Transaction on Pattern Analysis and Machine
Network Based Corner Detection Method, in: Proc
Intelligence, vol 12, no 9, pp 913-917, September
IEEE International Conference on Neural Networks,
1990.
Perth, WA, Australia, vol 4, pp 2116-2120,
18. S T Liu & W H Tsai, Moment-Preserving Corner November-December 1995.
Detection, Pattern Recognition, vol 23, no 5, pp 441-
31. M J J Wang , W Y Wu, L K Huang & D M Wang,
460, 1990.
Corner Detection using Bending Value, Pattern
19. J Cooper, S Venkatesh & L Kitchen, Early Jump-out Recognition Letters, vol 16, pp 575-583, 1995.
Corner Detectors, IEEE Transaction on Pattern
32. K K Lai & P S Y Wu, Effective Edge-Corner Detection
Analysis and Machine Intelligence, vol 15, no 8, pp
Method for Defected Images, in: Proc of 3rd
823-828, August 1993.
International Conference on Signal Processing,
20. C M Orange & F C A Greon, Model Based Corner Beijing, China, vol 2, pp 1151-1154, October 1996.
Detection, in: Proc IEEE Computer Society
33. H T Sheu & W C Hu, A Rotationally Invariant Two-
Conference on Computer Vision and Pattern
Phase Scheme for Corner Detection, Pattern
Recognition, New York, NY, USA, pp 690-691, June
Recognition, vol 29, no 5, pp 819-828, 1996.
1993.
34. F Arrebola, A Bandera, P Camacho & F Sandoval,
21. X Xie, R Sudhakar and H Zhuang, Corner Detection
Corner Detection by Local Histograms of Contour
by a Cost Minimization Approach, Pattern
Chain Code, Electronics Letters, vol 33, no 21, pp
Recognition, vol 26, no 8, pp 1235-1243, 1993.
1769-1771, 9th October, 1997.
22. S C Pei & J H Horng, Corner Point Detection using
35. F Arrebola, P Camacho, A Bandera and F Sandoval,
Nest Moving Average, Pattern Recognition, vol 27,
Corner Detection and Curve Representation by
no 11, pp 1533-1537, 1994.
Circular Histograms of Contour Chain Code,
23. U Seeger & R Seeger, Fast Corner Detection in Electronics Letters, vol 35, no 13, pp 1065-1067,
Gray-level Images, Pattern Recognition Letters, vol 24th June, 1999.
15, pp 669-675, 1994.
36. Q Ji & R M Haralick, Corner Detection with
24. K Sugiamoto & F Tomita, Boundary Segmentation by Covariance Propagation, in: Proc of Conference on
Detection of Corner, Inflection and Transition Points, Computer Vision and Pattern Recognition, San Juan,
in Proc IEEE Workshop on Visualization and Machine Puerto Rico, pp 362-367, June 1997.
Vision, Seattle, CA, USA, pp 13-17, June 1994.
37. B K Ray & K S Ray, Scale-Space Analysis and
25. X Zhang, R M Haralick & V Ramesh, Corner Corner Detection on Digital Curves using a Discrete
Detection using the MAP Technique, Conference A: Scale-Space Kernel, Pattern Recognition, vol 30, no
Computer Vision & Image Processing, Proceedings 9, pp 1463-1474, 1997.
38. D M Tsai, Boundary-Based Corner Detection using 51. M Marji & P Siy, A New Algorithm for Dominant
Neural Networks, Pattern Recognition, vol 30, no 1, Points Detection and Polygonization of Digital Curves,
pp 85-97, 1997. Pattern Recognition, vol 36, pp 2239-2251, 2003.
39. Z Zheng & D Zhao, Parallel Algorithm for Detecting 52. W Y Wu, Dominant Point Detection using Adaptive
Dominant Points on Multiple Digital Curves, Pattern Bending Value, Image and Vision Computing, vol 21,
Recognition, vol 30, no 2, pp 239-244, 1997. pp 517-525, 2003.
40. E Shilat, M Werman & Y Gdalyahu, Ridge’s Corner 53. C Urdiales, C Trazegnies, A Bandera & E Sandoval,
Detection and Correspondence, in: Proc IEEE Corner Detection Based on Adaptively Filtered
Conference on Computer Vision and Pattern Curvature Function, Electronics Letters, vol 39 no 5,
Recognition, San Juan, Puerto Rico, pp 976-981, pp 426-428, 6th March 2003.
June 1997.
54. M Banerjee, M K Kundu & P Mitra, Corner Detection
41. F Mokhtarian & R Suomela, Robust Image Corner using Support Vector Machines, in: Proceedings of
Detection through Curvature Scale Space, IEEE the 17th International Conference on Pattern
Transaction on Pattern Analysis and Machine Recognition, vol 2, pp 819-822, August 2004.
Intelligence, vol 20, no 12, pp 1376-1381, December
55. D S Guru, R Dinesh & P Nagabhushan, Boundary
1998.
Based Corner Detection and Localization using New
42. K Sohn, J H Kim & W E Alexander, A Mean Field ‘Cornerity’ Index: A Robust Approach, in:
Annealing Approach to Robust Corner Detection, Proceedings of the First Canadian Conference on
IEEE Transaction on Systems, Man and Cybernetics Computer and Robot Vision, pp 423-427, May 2004.
- Part B: Cybernetics, vol 28, no 1, pp 82-90,
56. F Arrebola & F Sandoval, Corner Detection and Curve
February 1998.
Segmentation by Multiresolution Chain-Code Linking,
43. L Li & W Chen, Corner Detection and Interpretation Pattern Recognition, vol 38, pp 1596-1614, 2005.
on Planar Curves using Fuzzy Reasoning, IEEE
57. M Sarfraz, A Rasheed, & Z Muzaffar, A Novel Linear
Transaction on Pattern Analysis and Machine
Time Corner Detection Algorithm, in: Proceedings of
Intelligence, vol 21, no 12, pp 1204-1210, December
the Computer Graphics, Imaging and Vision: New
1999.
Trends, pp 191-196, July 2005.
44. B Luo, A.D.J Cross & E.R Hancock, Corner
58. G Olague & B Hernandez, A New Accurate and
Detection Via Topographic Analysis of Vector-
Flexible Model Based Multi-Corner Detector for
Potential, Pattern Recognition Letters, vol 20, pp
Measurement and Recognition, Pattern Recognition
635-650, 1999.
Letters, vol 26, pp 27-41, 2005.
45. D M Tsai, H T Hou & H J Su, Boundary-Based
59. A C Poyato, N L F Garcýa, R M Carnicer & F.J M
Corner Detection using Eigen Values of Covariance
Cuevas, Dominant Point Detection: A New Proposal,
Matrices, Pattern Recognition Letters, vol 20, pp 31-
Image and Vision Computing, vol 23, pp 1226-1236,
40, 1999.
2005.
46. R M Guest & M C Fairhurst, A Clustering Approach
60. A Sobania & J P O Evans, Morphological Corner
to Corner Point Analysis in Hand Drawn Images, in:
Detector using Paired Triangular Structuring
Proceedings of the 16th International Conference on
Elements, Pattern Recognition, vol 38, pp 1087-1098,
Pattern Recognition, vol 3, pp 940-943, 2002.
2005.
47. N Ludtke, B Luo, E Hancock & R C Wilson, Corner
61. J S Lee, Y N Sun & C H Chen, Gray-Level-Based
Detection using a Mixture Model of Edge Orientation,
Corner Detection by using Wavelet Transform, IEEE
in: Proceedings of the 16th International Conference
TENCON, Beijing, China, pp 970-973, 1993.
on Pattern Recognition, vol 2, pp 574-577, 2002.
62. C H Chen, J S Lee & Y N Sun, Wavelet
48. H H Oh & S I Chien, Exact Corner Location using
Transformation for Gray-Level Corner Detection,
Attentional Generalized Symmetry Transform,
Pattern Recognition, vol 28, no 6, pp 853-861, 1995.
Pattern Recognition Letters, vol 23, pp 1361-1372,
2002. 63. J S Lee, Y N Sun & C H Chen, Multiscale Corner
Detection by Wavelet Transform, IEEE Transaction
49. F Shen & H Wang, Corner Detection Based on
on Image Processing, vol 4, no 1, pp 100-104,
Modified Hough Transform, Pattern Recognition
January 1995.
Letters, vol 23, pp 1039-1049, 2002.
64. K Kohlmann, Corner Detection in Natural Images
50. B K Ray & R Pandyan, ACORD – An Adaptive
Based on the 2-D Hilbert Transform, Signal
Corner Detector for Planar Curves, Pattern
Processing, vol 48, pp 225-234, 1996.
Recognition, vol 36, pp 703-708, 2003.
65. A Quddus & M M Falmy, An Improved Wavelet- and Tracking of Point Features, Technical Report,
Based Corner Detection Technique, in: Proc IEEE CMU-CS-91-132, 1991.
International Conference on Acoustics, Speech, and
80. B Lucas & T Kanade, An Iterative Registration
Signal Processing, Phoenix, AZ, USA, vol 6, pp 3213-
Technique with an Application to Stereo Vision, in:
3216, 1999.
Proc 7th International Joint Conference of Artificial
66. A Quddus & M M Falmy, Detection of Corners and Intelligence, vol 2, pp 674-679, 1981.
Smooth Joins using Wavelet Transform, in: Proc of 81. A Noble, Finding comers, Image and Vision
the 1999 IEEE International Symposium on Circuits Computing, vol 6, no 2, pp 121-128, 1988.
and Systems, Orlando, FL, USA, vol 3, pp 395-398,
1999. 82. R Deriche & G Giraudon, A Computational Approach
for Corner and Vertex Detection, International Journal
67. X Peng, C Zhou & M Ding, Corner Detection Method of Computer Vision, vol 10, no 2, pp 101-124, 1993.
Based on Wavelet Transform, in: Proc SPIE, vol
4550, pp 319-323, 2001. 83. J Shi & C Tomasi, Good Features to Track, in: Proc
IEEE Conference on Computer Vision and Pattern
68. X Gao, F Sattar, A Quddus & R Venkateswarlu, Recognition, Seattle, CA, USA, pp 593-600, June
Corner Detection of Contour Images using 1994.
Continuous Wavelet Transform, ICICS-PCM,
Singapore, pp 724-728, 15-18 December, 2003. 84. H Wang, and M Brady, A Practical Solution to Corner
Detection, in: Proc 5th International Conference on
69. C H Yeh, Wavelet-Based Corner Detection using Image Processing, Austin, TX, USA, vol 1, pp 919-
Eigen Vectors of Covariance Matrices, Pattern 923, November 1994.
Recognition Letters, vol 24, pp 2797- 2806, 2003.
85. H Wang, and M Brady, Real-Time Corner Detection
70. L Sun, Y.Y Tang & X You, Corner Detection for Algorithm for Motion Estimation, Image and Vision
Object Recognition by using Wavelet Transform, Computing, vol 13, no 9, pp 695-703, 1995.
Electronics Letters, vol 38 no 14, pp 699-701, July
86. Y Cui & P D Lawrence, Detecting Scale-Space
2002
Consistent Corners Based on Corner Attributes,
71. H P Moravec, Towards Automatic Visual Obstacle IEEE International Conference on Systems, Man and
Avoidance, in: Proc 5th International Joint Conference Cybernetics, ‘Intelligent Systems for the 21st
of Artificial Intelligence, pp 584, Cambridge, Century’, Vancouver, BC, Canada, vol 4, pp 3549-
Massachusetts, 1977. 3554, 1995.
72. P R Beaudet, Rotationally Invariant Image Operators, 87. K J Lee & Z Bien, A Gray-Level Corner Detector
in: Proc International Joint Conference of Pattern using Fuzzy Logic, Pattern Recognition Letters, vol
Recognition, pp 579-583, Kyoto, 1978. 17, pp 939-950, 1996.
73. L Kitchen & A Rosenfeld, Gray Level Corner 88. M E Dýaz, J Domingo & G Ayala, A Gray-Level 2D
Detection, Pattern Recognition Letters, vol 1, no 2, Feature Detector using Circular Statistics, Pattern
pp 95-102, 1982. Recognition Letters, vol 18, pp 1083-1089, 1997.
74. Z O Wu & A Rosenfeld, Filtered Projection as an Aid 89. S M Smith & J M Brady, A New Approach to Low-
to Corner Detection, Pattern Recognition, vol 16, no Level Image Processing, International Journal of
31, pp , 1983. Computer Vision, vol 23, no 1, pp 45-78, 1997.
90. J Kautsky, B Zitova, J Flusser & G Peters, Feature
75. O A Zuniga & R M Haralick, Corner Detection using
Point Detection in Blurred Images, in: Proc
the Facet Model, in: Proc Conference on Computer
International Conference IVCNZ ’98, Image and Vision
Vision and Pattern Recognition, pp 30-37, 1983.
Computing Proceedings (Klette R., Gimelfarb G.,
76. K Paler, J Foglein, J Illingworth & J Kittler, Local Kakarala R eds.), University of Auckland, Auckland,
Ordered Gray-Levels as an Aid to Corner Detection, New Zealand, pp 103-108, November 1998.
Pattern Recognition, vol 17, no 5, pp 535-543, 1984.
91. B Zitova, J Kautsky, G Peters & J Flusser, Robust
77. C Harris, and M Stephens, “A Combined Corner and Detection of Significant Points in Multiframe Images,
Edge Detector”, 4th Alvey Vision Conference, pp Pattern Recognition Letters, vol 20, pp 199-206,
147-151, Manchester, UK, 1988. 1999.
78. W Forstner & E Gulch, A Fast Operator for Detection 92. R Laganiere, A Morphological Operator for Corner
and Precise Location of Distinct Points, Corners and Detection, Pattern Recognition, vol 31, no 11, pp
Centers of Circular Features, ISPRS Inter 1643-1652, 1998.
Commission Workshop, pp 147-151, June 1987. 93. R S Lin, C H Chu & Y C Hsueh, A Modified
79. C Tomasi & T Kanade, Shape and Motion from Image Morphological Corner Detector, Pattern Recognition
Streams: A Factorization Method-Part 3, Detection Letters, vol 19, pp 279-286, 1998.
94. T Stammberger, M Michaelis, M Reiser & K H Canadian Conference on Electrical & Computer
Englmeier, A Hierarchical Filter Scheme for Efficient Engineering, pp 912-916, 2002.
Corner Detection, Pattern Recognition Letters, vol
108. I Golightly & D Jones, Corner Detection and Matching
19, pp 687-700, 1998.
for Visual Tracking during Power Line Inspection,
95. M Trajkovic & M Hedley, Fast Corner Detection, Image and Vision Computing, vol 21, pp 827-840,
Image and Vision Computing, vol 16, no 2, pp 75-87, 2003.
1998.
109. K Mikolajczyk & C Schmid, Scale and Affine Invariant
96. F Chabat, G Z Yang & D M Hansell, A Corner Interest Point Detectors, International Journal of
Orientation Detector, Image and Vision Computing, Computer Vision, vol 60, no 1, pp 63-86, 2004.
vol 17, pp 761-769, 1999.
110. S Alkaabi & F Deravi, Candidate Pruning for Fast
97. X Lv & J Zhou, Robust Corner Detection under Corner Detection, Electronic Letters, vol 40, Issue 1,
Varying Illumination, in: Proc International Conference pp 18-19, 2004.
on Information Intelligence and Systems, Bethesda,
MD, USA, pp 396-398, 1999. 111. D Zhou, Y H Liut & X Cai, An Efficient and Robust
Corner Detection Algorithm, in: Proc 5th World
98. E Sojka, A New Algorithm for Detecting Corners in Congress on Intelligent Control and Automation,
Digital Images, in: Proc of the 18th Spring Conference Hangzhou P.R China, pp 4020-4024, June 2004.
on Computer Graphics, Budmerice, Slovakia, pp 55-
62, 2002. 112. C S Kenney, M Zuliani & B S Manjunath, An Axiomatic
Approach to Corner Detection, in: Proc of the IEEE
99. M A Ruzon & C Tomasi, Corner Detection in
Computer Society Conference on Computer Vision
Textured Color Images, in: Proc Seventh IEEE
and Pattern Recognition, San Diego, CA, USA, pp
International Conference on Computer Vision,
191-197, June 2005.
Kerkyra, Greece, vol 2, pp 1039-1045, 1999.
113. E Vincent & R Laganiere, Detecting and Matching
100. J Basak & D Mahata, A Connectionist Model for
Feature Points, Journal of Visual Communication and
Corner Detection in Binary and Gray Images, IEEE
Image Representation, vol 16, pp 38-54, 2005.
Transaction on Neural Networks, vol 11, no 5, pp
1124-1132, September 2000. 114. T Cooke & R Whatmough, Using Learning Algorithms
101. F Deschenes & D Ziou, Detection of Line Junctions to Improve Corner Detection, in: Proc Digital Image
and Line Terminations Using Curvilinear Features, Computing: Technqiues and Applications, pp 371-
Pattern Recognition Letters, vol 21, pp 637-649, 378, December 2005.
2000. 115. S C Pei & J J Ding, New Corner Detection Algorithm
102. L Alvarez, C Cuenca & L Mazorra, Morphological by Tangent and Vertical Axes and Case Table, in:
Corner Detection Application to Camera Calibration, Proc IEEE International Conference on Image
in: Proceedings of the IASTED International Processing, vol 1, pp 365 – 368, September 2005.
Conference Signal Processing, Pattern Recognition 116. E Rosten & T Drummond, Machine Learning for High-
& Applications, Rhodes, Greece, pp 21-26, July 2001. Speed Corner Detection, in: Proc European
103. R.P Wurtz & T Lourens, Corner Detection in Color Cinference on Computer Vision, pp 430-443, May
Images through a Multiscale Combination of End- 2006.
Stopped Cortical Cells, Image and Vision Computing, 117. A Quddus & M M Falmy, Corner Detection using
vol 18, pp 531-541, 2000. Gabor-type Filtering, in: Proceedings of the IEEE
104. B Telle & M J Aldon, Interest Points Detection in International Symposium on Circuits and Systems,
Color Images, in: Proc IARP Workshop on Machine vol 4, Monterey, CA, USA, pp 150-153, June 1998.
Vision Applications, Nara (Japan), pp 550-553,
118. F Pedersini, E Pozzoli, A Sarti & S Tubaro, Multi-
December 11-13, 2002
Resolution Corner Detection, in: Proceedings of the
105. X Gao, Z Yu, F Sattar & R Venkateswarlu, Multi International Conference on Image Processing, vol
Scale Corner Detection for Gray Level Images using 3, Vancouver, BC, Canada, pp 881-884, 2000.
Plessey Method, in: Proc 8th International
119. A Quddus & M Gabbouj, Wavelet-Based Corner
Conference on Control, Automation, Robotics and
Detection Algorithm Technique using Optimal Scale,
Vision, Kunming, China, pp 363-368, December 2004.
Pattern Recognition Letters, vol 23, pp 215-220,
106. S Bae, I S, Kweon & C D Yoo, COP: A New Corner 2002.
Detector, Pattern Recognition Letters, vol 23, no 11,
pp 1349-1360, 2002. 120. X Gao, F Sattar & R Venkateswarlu, Corner
Detection of Gray Level Images using Gabor
107. K Elias & R Laganiere, Cones: A New Approach Wavelets, in: Proc International Conference on Image
Towards Corner Detection, in: Proc of the IEEE Processing, vol 4, pp 2669-2672, October 2004.
Authors
Ambar Dutta born on 3rd December, 1977, did his BSc (Honors) in Mathematics from
Presidency College, Kolkata in 1999 and MCA from Jadavpur University, Kolkata in 2002. He is
working as a Lecturer in the department of Computer Science and Engineering, Birla Institute of
Technology, Mesra, Kolkata Extension Centre. He is pursuing his PhD from Jadavpur
University, Kolkata in the area of image processing (corner detection and matching). His
research interest includes Image Processing, Pattern Recognition, Data Mining and Soft
Computing.
Address: Department of Computer Science and Engineering, Birla Institute of Technology,
Mesra, Kolkata Extension Centre, Southend Conclave, 1582, Rajdanga Main Road,
Kolkata 700 107, India.
email: <adutta@bitmesra.ac.in>
* * *
Avijit Kar did his MSc and PhD in 1980 and 1984 respectively from IIT Kharagpur. He is
currently a professor in the department of Computer Science and Engineering in Jadavpur
University, Kolkata. He has supervised several PhD theses and is actively involved in many R
& D activities and IT related consultancy for Government of India and the private sector. His
research interest includes biomedical imaging as well as SAR imaging. He is also into computer
systems reliability. He is involved in a large number of industry sponsored development
projects.
Address: Department of Computer Science and Engineering, Jadavpur University, Kolkata
700 032, India.
email: <avijit_kar@cse.jdvu.ac.in>
* * *
B N Chatterji born on 10th November, 1942, obtained BTech (Hons) (1965) and PhD (1970) in
Electronics and Electrical Communication Engineering of IIT, Kharagpur. He did Post Doctoral
work at University of Erlangen-Nurenberg, Germany during 1972-73. Worked with Telerad Pvt
Ltd, Bombay (1965), Central Electronics Research Institute, Pilani (1966) and IIT, Kharagpur as
faculty member during 1967-2005. He was Professor during 1980-2005, Head of the Department
during 1987-1991, Dean Academic Affairs during 1994-1997 and Member of Board of Governors
of IIT, Kharagpur during 1998-2000. He has published about 150 journal papers, 200 conference
papers and four books. He was Chairman of four International Conferences and ten National
Conferences. He has coordinated 25 short-term courses and was the chief investigator of 24
Sponsored Projects. He is the Fellow/Life Member/Member of eight Professional Societies. He
has received ten National Awards on the basis of his Academic/Research contributions. His
areas of interests are Pattern Recognition, Image Processing, Signal Processing, Parallel
Processing and Control Systems.
Address: Department of Electronics Communication Engineering, B P Poddar Institute of
Management and Technology, 137, VIP Road, Kolkata 700 052, India.
email: <bnchatterji@gmail.com>
ABSTRACT
Uncompressed multimedia data (graphics, audio and video) requires considerable storage and transmission bandwidth.
A fundamental goal of data compression is to reduce the bit rate for transmission or storage while maintaining
acceptable quality. Image coding and compression technique, converts an image that requires low memory storages
space, smaller bandwidth for transmission, high PSNR with acceptable image quality. This paper presents wavelet
based coding algorithm SPIHT to encode and compress an image data. The coding and decoding process is
comparatively fast. The numerical results obtained using MATLAB shows that the output image has high value of
PSNR with good compression ratio for low bit rate.
functions. These sets are divided into four parts image component & HH component is the result of
such as approximation, horizontal details, vertical high pass filtering of rows & columns [5].
details and diagonal details; these components are
called as nodes of a tree as shown in Fig 1 [3]. 2 THE PEAK SIGNAL TO NOISE
RATIO (PSNR)
N–1
LH2 HH2
⎢ ⎢2
MSE = Σ ⎢ xi – x^i ⎢ (2)
i=0 ⎢ ⎢
LH1 HH1 ⎛ M 2 ⎞
PSNR = 10 log10 ⎜ ⎯⎯⎯⎯ ⎟ (3)
⎝ MSE ⎠
Where xi , x^i are the input & reconstructed pixel
values in the image respectively & M is the
maximum peak to peak value in the image (typically
256 for 8 bit image). A good PSNR performance is
Fig 1 Pyramidal structure of 3-level wavelet a prerequisite for any modern compression
decomposition algorithm [7,8].
decreasing thresholds, with the initial threshold all the coefficients in the current set (which initially
being the largest power of 2, which is smaller than are full subtrees of descendants). When all the
the magnitude of the largest coefficient. At any coefficients in a set are insignificant, that fact can
time, coefficients with magnitudes larger than the be indicated by outputting only one bit. When a
current threshold are deemed to be significant and given set contains a significant pixel, it is partitioned
others insignificant. The coefficients are considered into subsets (subtrees rooted at the descendants of
to be organized into trees, as shown in Fig 2. the current root) and those in turn are tested for
Normally, most of the image energy is concentrated significance. As before, when significant pixels are
in the low frequency components. Consequently, identified, their sign is transmitted and they are
the variance decreases as we move from the moved to the list of significant pixels. After a pass
highest to the lowest levels of the wavelet pyramid. has been completed through the lists of insignificant
It has been observed that there is a spatial self- sets and pixels, a ‘refinement’ pass is made through
similarity between levels, and the coefficients are the list of significant pixels (not including those just
expected to be magnitude ordered if we move added in the last sorting pass) and the values of
downward in the pyramid following the same spatial those coefficients are refined by one bit. After an
orientation [7]. entire pass is made, the threshold is decreased by
a factor of two and another sorting pass is initiated.
This organization will allow us to identify large
subtrees as containing no significant pixels. This The output bitstream thus consists of the result
algorithm maintains lists of insignificant sets, of significance tests and signs and refinement bits
insignificant pixels, and significant pixels, and is of currently significant coefficients. The location of
initialized with the list of insignificant pixels being the coefficients being refined or classified as
the nodes in the highest layer and the list of significant is never explicitly transmitted; it is
insignificant sets being the sets of the subtree known implicitly since the encoder and decoder
descendants of each such node. In a ‘sorting’ pass, both share the same algorithm, and since all
the algorithm works its way down the list of branching decisions as the encoder searches the
insignificant pixels first, testing their magnitude coefficients are output. Arithmetic coding can be
against the current threshold, outputting their applied to the output bitstream to further compress
significance, and, when one is significant, the results of the significance tests, at the expense
outputting its sign and moving it to the list of of more computation time. The signs and refinement
significant pixels. Then, it moves through the list of bits both tend to be 0 or 1 with equal probability, so
insignificant sets, performing the magnitude test to there is nothing to gain by compressing them.
Proceeding in this manner, the algorithm transmission, which always selects the most
transmits approximations of the most important important information that yields the largest
coefficients (those with largest magnitude) first, distortion reduction to transmit next. If the wavelet
and refines the values of all significant coefficients transform is unitary, then the Euclidean norm is
one bit at a time [1]. This results in a progressive preserved and it can be shown that such a
Image MRI (256X256) & Lena (512X512) & Cameramen (256X256) Rice (256X256) &
8 bit 8 bit &12 bit 12 bit
Bit Rate PSNR Compression PSNR Compression PSNR Compression PSNR Compression
(Bits / Sec) (dB) Ratio (%) (dB) Ratio (%) (dB) Ratio (%) (dB) Ratio (%)
transmission scheme is the optimal way to MATLAB codes are tested with images of different
decrease the RMS error in the reconstructed image size & type. From the results it has been found that
[6]. The bitstream, which is generated, is fully increasing bit rate, PSNR increases & compression
embedded, which means that it can be truncated (or ratio decreases. The results of this coding algorithm
the compression process stopped) at any point and with its embedded code and reduced execution
the image decompressed and reconstructed. In time are so impressive that it can be used for
fact a single file compressed at a high bit rate (low standardization in many image compression
compression ratio) can be decompressed at any systems.
smaller bit rate (higher compression ratio) and the
results will be identical to having compressed the 6. REFERENCES
file at the higher ratio to begin with. The algorithm
achieves fully progressive transmission: at any 1. A Said & W A Pearlman, Image compression using
the special orientation tree, IEEE Int Symp Circuits
time the transmission of another byte resulting in
and System, Chicago, pp 279-282, May 1993.
simply further refining the values of one or more
coefficients. Thus, the desired compression ratio 2. A Said & W A Pearlman, An Image Multiresolution
Representation for lossless & lossy Image
or bit rate can be fully specified in advance in
compression, IEEE transaction on image processing,
contrast to JPEG or to the conventional wavelet vol 15, pp 1303-1310, Sept 1996.
approach.
3. A Manduca & A Said, Wavelet Compression of
medical images with SPIHT, SPIE symposium on
4. NUMERICAL RESULTS Medical imaging, Cambridge, MA, March 1996.
4. R A Dvorer, B Jawerth & B J Lucier, Image
compression through wavelet transform coding, IEEE
The SPIHT coding/decoding algorithm has been Transactions on Information Theory, vol 38, pp 719-
implemented in MATLAB. It has been tested on 746, Mar 1992.
images of different of size & type. The DWT level is 5. P N Topiwala, Wavelet Image & Video Compression,
kept equal to 3. The numerical results are presented Kluwer Academic publisher 2002.
in tabular form (Table 1) in terms of PSNR values &
6. K P Soman & K I Ramchandran, Insight into
Compression ratio for different bit rate. Figure 5 wavelets, from theory to practice, PHI, New Delhi
shows the comparative performance of the 2004.
proposed algorithm. 7. M Vetterli J Kovaccevic, Wavelets & Subband
Coding, Prentice Hall PTR, Englewood cliffs, NJ.
5. CONCLUSION 8. Yong Sun, Hui Zhang & Guangshu Hu, Real-time
implementation of a new low-memory SPIHT image
In this paper wavelet based SPIHT image coding algorithm using DSP chip, IEEE Transactions
coding algorithm has been presented. The designed on Image Processing 11(9), pp 1112-1116, 2002.
Authors
N B Chopade, received MSc (Applied Electronics) & ME (Electronics) degrees from SGB
Amravati University, Amravati in the year 1992 & 1998 respectively. He scored third merit
position in Amravati University in the year 1992. His areas of research are DSP, Image
processing, Wavelets Applications. Currently he is pursuing PhD from SGB Amravati University,
Amravati. He has joined SSGM College of Engineering, Shegaon in 1992 and is currently
^
Address: Dr. Babasaheb Ambedkar Technological University, Lonere, Dist Raigad 402 103,
India.
email: <vc_2005@rediffmail.com>
Subject of on-line submission & review of articles for IETE Journals has been under
consideration for a long time. Keeping the present trend of Internet, automation and
computerization at different levels, it was considered prudent and necessary to move from
manual hard copy submission and review system to on-line system. The proposed software
is expected to help authors, Honorary Editors, Reviewers, Editors-in-Chief and IETE HQs
to streamline submission and reviewing process for timely publication of journals.
The procurement of software system has been approved by the Council of IETE and will
be operationalized in the near future by end September 2008. Effective date for system
will be publicised in IETE Journals and website.
The salient features of the proposed system are as follows:-
a) The system will create the website and a separate button to choose the journal from
the list. All the three journals will appear on the page. The authors will choose one of
the journal and upload the manuscript. These will be menu driven and simple.
b) Resource templates based on approved standard would be hosted on IETE website
along with the guidelines to authors for submission of articles on line with domain
address www.ietejournals.org.
c) The chairpersons of the editorial boards would be the Editors-in-Chief for the three
journals, i.e. IETE Journal of Education (JE), IETE Journal of Research (JR) and IETE
Technical Review (TR).
d) Each journal will have area editors from across the world as empanelled by the
respective editor-in-chief. These panels would be periodically reviewed taking into
consideration the performance of the panelists.
e) The concerned Editor-in-Chief will carry out on-line preliminary review for further
processing or out-right rejection with intimation to author. The IETE HQ will keep the
record of entire submission, revision, rejection and acceptance details.
f) After final acceptance by the Editor-in-Chief, an e-mail will be generated automatically
to the Administrator, Publication Section at IETE HQs who will have the access to the
word / pdf document of the final accepted manuscript. The administrator will download
it and carry out all editorial changes before it is given to Press for printing of the journals.
h) Authors are requested not to submit any articles to IETE, neither through e-mail
nor by post, till the on-line system is available, which is expected by end
September 2008.
g) The printed journals will be hosted on IETE website for downloading by readers / authors
and will be available on IETE website.
Any suggestions and feedback are most welcome. Please email to ietend@gmail.com
Registered with the Registrar of Newspapers for India, Delhi under Regn. No. 42814/84
Printed and Published by Brig (Retd) V K Panday for the Institution of Electronics and Telecommunication Engineers,
2, Institutional Area, Lodi Road, New Delhi 110 003 (India) and printed at the Shivam Offset Press, A-12/1, Naraina Indl. Area,
Phase I, New Delhi-110 028 (India). Dy Managing Editor : CDR A P Sharma, IN (Retd) • Copyright © 2008 by the IETE