1

A Project Report
On
FINGERPRINT RECOGNITION
Submitted to JNTU in partial fulfillment of the requirement for the award of Bachelor

of Technology
In

Electronics and Communication Engineering

By
HEENA TARANUM

07361A0430

SWATHI

07361A04A2

CHAITANYA KUMARI

07361A0416

Under the valuable guidance of

Prof. SANDEEP V.M.

Department of
ELECTRONICS AND COMMUNICATION ENGINEERING
JAYA PRAKASH NARAYAN COLLEGE OF ENGINEERING
DHARMAPUR, MAHABUBNAGAR-509001(A.P.)

AFFILIATED TO JNTU, ACCREDITED BY N.B.A

2010-2011
1

DEPARTMENT OF
ELECTRONICS & COMMUNICATION ENGINEERING

JAYAPRAKASH NARAYAN COLLEGE OF

ENGINEERING
DHARMAPUR, MAHABOOBNAGAR 509381 (A.P.)
Affiliated to J.N.T.U., Accredited by N.B.A

CERTIFICATE
This is to certify that the Project on FINGERPRINT
RECOGNITION is a bonafide work done by Mr. /Ms. HEENA
TARANUM bearing Roll. No. 07361A0430 in partial fulfillment of the
requirement of the award for the degree of Bachelor of Technology in
Electronics and Communications Engineering J.N.T.U., Hyderabad during
the year 2010-2011.

Guide and H.O.D.

Sri. Prof.Sandeep V.M.
Professor& Head,

External Examiner:

Dept of E.C.E.

INDEX

INDEX
1. Abstract
2. Introduction
3. Background
3.1. What is fingerprint?
3.2. Classifying fingerprints
3.3. History of fingerprint recognition
3.3.1. William west
3.3.2. Minutia-based algorithm
3.3.3. Spectrum analysis
3.3.4. Hybrid technology

4. Fingerprint recognition
4.1. Preprocessing
4.1.1. Image enhancement
4.1.2. Noise reduction
4.1.3. Binarization
4.1.4. Thinning

4.2. Segmentation
4.3. Feature extraction

4.3.1. Global extraction

4.3.1.1. Mean and STD of binary image
4.3.1.2. Mean and STD of gray image

4.3.2. Local extraction

4.3.2.1. Level 1
4.3.2.2. Level 2
4.3.3. Data base
4.3.3.1. Average of images, then finding feature
4.3.3.2. Finding feature, then averaging values

5. Classification of scattering
6. Conclusion and future scope
7. References

ABSTRACT

1. ABSTRACT
A fingerprint is an imprint made by the pattern of ridges on the pad of a human finger, in our
process; segmentation is takes place, is the goal of segmentation is to simplify and/or change
the representation of an image into something that is more meaningful and easier to analyze.
For feature extraction, We are using global method and local method. Many object
recognition systems use global features that describe an entire image. Global method can be
done in two ways. One way is for binary image, another is for gray image.
Local feature is an image pattern which differs from its immediate neighborhood. It is usually
associated with a change of image property or several properties simultaneously. The image
properties considered as feature. Local method can be done in two ways. One way is level 1,
another is level 2.

INTRODUCTION

2. INTRODUCTION
Identity verification in computer systems is done based on measures like keys, cards,
passwords, PIN and so forth. Unfortunately, these may often be forgotten, disclosed or
changed. A reliable and accurate identification/verification technique may be designed using
biometric technologies, which are further based on the special characteristics of the person
such as face, iris, fingerprint, signature and so forth. This technique of identification is
preferred over traditional passwords and PIN-based techniques for various reasons:

The person to be identified is required to be physically present at the time of

identification.

Identification based on biometric techniques obviates the need to remember a

password or carry a token.

A biometric system essentially is a pattern recognition system that makes a personal

identification by determining the authenticity of a specific physiological or behavioral
characteristic possessed by the user. Biometric technologies are thus defined as the
automated methods of identifying or authenticating the identity of a living person based on a
physiological or behavioral characteristic. A biometric system can be either an identification
system or a verification (authentication) system; both are defined below.

Identification: One to Many A comparison of an individuals submitted biometric

sample against the entire database of biometric reference templates to determine
whether it matches any of the templates.

Verification: One to One A comparison of two sets of biometrics to determine if

they are from the same individual.

Biometric authentication requires comparing a registered or enrolled biometric sample

(biometric template or identifier) against a newly captured biometric sample (for example,
the one captured during a login). This is a three-step process (Capture, Process, Enroll)
followed by a Verification or Identification.

10

During Capture, raw biometric is captured by a sensing device, such as a fingerprint scanner
or video camera; then, distinguishing characteristics are extracted from the raw biometric
sample and converted into a processed biometric identifier record (biometric template). Next
is enrollment, in which the processed sample (a mathematical representation of the template)
is stored/ registered in a storage medium for comparison during authentication. In many
commercial applications, only the processed biometric sample is stored. The original
biometric sample cannot be reconstructed from this identifier.
Among the various biometric technologies being considered are fingerprint, facial features,
hand geometry, voice, iris, retina, vein patterns, palm print, DNA, keystroke dynamics, ear
shape, odor, signature and so forth.

2.1 Fingerprint
Fingerprint biometric is an automated digital version of the old ink-and-paper method used
for more than a century for identification, primarily by law enforcement agencies (Maltoni,
2003). The biometric device requires each user to place a finger on a plate for the print to be
read. Fingerprint biometrics currently has three main application areas: large-scale Automated
Finger Imaging Systems (AFIS), generally used for law enforcement purposes; fraud
prevention in entitlement programs; and physical and computer access.
A major advantage of finger imaging is the long-time use of fingerprints and its wide
acceptance by the public and law enforcement communities as a reliable means of human
recognition. Others include the need for physical contact with the optical scanner, possibility
of poor-quality images due to residue on the finger such as dirt and body oils (which can
build up on the glass plate), as well as eroded fingerprints from scrapes, years of heavy labor
or mutilation.

2.2 fingerprint recognition

In our method firstly we reducing the noise and then segmentation is done.
In our method we are using for feature extraction mainly two methods, they are global
method and local method. In global method we are finding mean and varience of binary
image and original image.
10

11

By using binary image (in this we converting original image into binary and proceeding next
steps), finding the images means and averaging all values to get a single value. This single
value is used for matching by comparing that value to each of the mean value of image. But
this method is providing better result so going for another method.
In another method instead of binary image we using original image, and calculating same
process as mentioned above. And matching is done by using that average value and
individual mean value of image value .in this process we find the better result than previous
one.
In local method, Local feature is an image pattern which differs from its immediate
neighborhood. In this method we are dividing into some equal windows in 3*3 or 5*5 or 7*7
window. This entire process is done into two types; they are level one and level2.
In level 1, only taking the mean and standard deviation of the image with respect to that
window. And taking the average of that individual values and comparison is done between
average value and individual value.
In level2 level1 process and mean of mean ,mean of std , std of mean and std of std is
calculating for individual images and taking average of that images (m of m, m of s, s of m
and s of s are averaging separately).at last we comparing with all image individual values to
average value.

11

12

BACK GROUND

12

13

3. BACK GROUND
3.1 WHAT IS FINGER PRINT?
3.1.1 Fingerprint:
1. The characteristic dermal ridges on the finger. This is the original meaning of fingerprint.
2. The characteristic pattern of the peptide fragments of a protein that have been subjected
electrophoresis and, at a right angle, chromatography. Peptide fingerprinting was invented by
Vernon Ingram in 1957.
3. The characteristic pattern of DNA fragments identified by Southern hybridization or by
PCR (polymerase chain reaction) DNA fingerprinting was invented by Alec Jeffreys in 1984.
A fingerprint is an imprint made by the pattern of ridges on the pad of a human finger.
These prints are often left on objects at a crime scene and are therefore used in forensic
science to identify suspects. No two humans have ever been found to have identical
fingerprints, thus fingerprints are widely believed to be unique.
Sometimes the prints are invisible, in which case they are called latent fingerprints, but there
are chemical techniques such as cyanoacrylate fuming and ninhydrin spray that can make
them visible.
Recently the American Federal Bureau of Investigation adopted a wavelet-based system for
efficient storage of fingerprint data, developed by Ingrid Daubechies.
In the 2000s, electronic fingerprint readers have been introduced for security applications
such as identification of computer users (log-in authentication). However, early devices have
been discovered to be vulnerable to quite simple methods of deception, such as fake
fingerprints cast in gels.
There is some controversy over the uniqueness of fingerprints. Even those who accept their
uniqueness sometimes argue that the techniques used to compare fingerprints are fallible. The
same fingerprint as it would be detected on a surface.

13

14

Fingerprint analysis (or Dactylographic, a term mainly used in the US) is the science of
using fingerprints to uniquely identify someone. Humans leave behind prints of the ridges of
the skin on their fingertips when handling certain materials. The pattern of ridges is thought
to be unique for each person and in practice has proved unique enough to identify the person
who left the fingerprint.

3.2 Classifying fingerprints:

The Galton-Henry system of fingerprint classification, published in June 1900, was A
fingerprint is the impression made by the papillary ridges on the ends of the fingers and
thumbs. Fingerprints afford an infallible means of personal identification, because the ridge
arrangement on every finger of every human being is unique and does not alter with growth
or age. Fingerprints serve to reveal an individual's true identity despite personal denial,
assumed names, or changes in personal appearance resulting from age, disease, plastic
surgery, or accident. The practice of utilizing fingerprints as a means of identification,
referred to as dactyloscopy, is an indispensable aid to modern law enforcement.
Each ridge of the epidermis (outer skin) is dotted with sweat pores for its entire length and is
anchored to the dermis (inner skin) by a double row of peg like protuberances, or papillae.
Injuries such as superficial burns, abrasions, or cuts do not affect the ridge structure or alter
the dermal papillae, and the original pattern is duplicated in any new skin that grows. An
injury that destroys the dermal papillae, however, will permanently obliterate the ridges.
Any ridged area of the hand or foot may be used as identification. However, finger
impressions are preferred to those from other parts of the body because they can be taken
with a minimum of time and effort, and the ridges in such impressions form patterns
(distinctive outlines or shapes) that can be readily sorted into groups for ease in filing.
Early anatomists described the ridges of the fingers, but interest in modern fingerprint
identification dates from 1880, when the British scientific journal Nature published letters by
the Englishmen Henry Faulds and William James Herschel describing the uniqueness and
permanence of fingerprints.
Their observations were experimentally verified by the English scientist Sir Francis Galton,
who suggested the first elementary system for classifying fingerprints based on grouping the
patterns into arches, loops, and whorls. Galton's system served as the basis for the fingerprint
14

15

classification systems developed by Sir Edward R. Henry, who later became chief
commissioner of the London metropolitan police, and by Juan Vucetich of Argentina.
Officially introduced at Scotland Yard in 1901 and quickly became the basis for its criminalidentification records. The system was adopted immediately by law-enforcement agencies in
the English-speaking countries of the world and is now the most widely used method of
fingerprint classification. Juan Vucetich, an employee of the police of the province of Buenos
Aires in 1888, devised an original system of fingerprint classification published in book form
under the title Dactiloscopa Compared (1904; "Comparative Fingerprinting"). His system is
still used in most Spanish-speaking countries.
Fingerprints are classified in a three-way process: by the shapes and contours of individual
patterns, by noting the finger positions of the pattern types, and by relative size, determined
by counting the ridges in loops and by tracing the ridges in whorls. The information obtained
in this way is incorporated in a concise formula, which is known as the individual's
fingerprint classification.
There are several variants of the Henry system, but that used by the Federal Bureau of
Investigation (FBI) in the United States recognizes eight different types of patterns: radial
loop, ulnar loop, double loop, central pocket loop, plain arch, tented arch, plain whorl, and
accidental. Whorls are usually circular or spiral in shape. Arches have a mound like contour,
while tented arches have a spike like or steeple like appearance in the center. Loops have
concentric hairpin or staple-shaped ridges and are described as "radial" or "ulnar" to denote
their slopes; ulnar loops slope toward the little finger side of the hand, radial loops toward the
thumb. Loops constitute about 65 percent of the total fingerprint patterns; whorls make up
about 30 percent and arches and tented arches together account for the other 5 percent. The
most common pattern is the ulnar loop.
Dactyloscopy, the technique of fingerprinting, involves cleaning the fingers in benzene or
ether, drying them, and then rolling the balls of each over a glass surface coated with printer's
ink. Each finger is then carefully rolled on prepared cards according to an exact technique
designed to obtain a light gray impression with clear spaces showing between each ridge so
that the ridges may be counted and traced. Simultaneous impressions are also taken of all
fingers and thumbs.

15

16

Latent fingerprinting involves locating, preserving, and identifying impressions left by a

culprit in the course of committing a crime. In latent fingerprints, the ridge structure is
reproduced not in ink on a record card but on an object in sweat, oily secretions, or other
substances naturally present on the culprit's fingers. Most latent prints are colorless and must
therefore be "developed," or made visible, before they can be preserved and compared. This
is done by brushing them with various gray or black powders containing chalk or lampblack
combined with other agents. The latent impressions are preserved as evidence either by
photography or by lifting powdered prints on the adhesive surfaces of tape.
Though the technique and its systematic use originated in Great Britain, fingerprinting was
developed to great usefulness in the United States, where in 1924 two large fingerprint
collections were consolidated to form the nucleus of the present file maintained by the
Identification Division of the FBI. The division's file contained the fingerprints of more than
90 million persons by the late 20th century. Fingerprint files and search techniques have been
computerized to enable much quicker comparison and identification of particular prints.
Other "fingerprinting" techniques have also been developed. These include the use of a sound
spectrograph--a device that depicts graphically such vocal variables as frequency, duration,
and intensity--to produce voice graphs, or voiceprints, and the use of a technique known as
DNA fingerprinting, an analysis of those regions of DNA that vary among individuals, to
identify physical evidence (blood, semen, hair, etc.) as belonging to a suspect. The latter test
has been used in paternity testing as well as in forensics.
There are three basic fingerprint patterns: Arch, Loop and Whorl. There are more complex
classification systems that further break down the pattern to plain arches or tented arches.
Loops may be radial or ulnar. Whorls also have smaller classifications. However, the five
most commonly used are: whorl, right loop, left loop, arch and tented arch.
There are three main fingerprint patterns: arches, loops and whorls.
Arches are found in about 5% of fingerprint patterns encountered. The ridges run from one
side to the other of the pattern, making no backward turn. Ordinarily, there is no delta in an
arch pattern but where there a delta; no re-curving ridge must intervene between the core and
delta points. There are four types of arch patterns: plain arches, radial arches, ulnar arches
and tented arches. Plain arches have an even flow of ridges from one side to the other of the
pattern; no significant up thrusts and the ridges enter on one side of the impression, and
16

17

flow out the other with a rise or wave in the center. The ridges of radial arches slope towards
the thumb, have one delta and no re-curving ridges. On ulnar arches, the ridges slope towards
the little finger, have one delta and no re-curving ridges. Tented arches have an angle, an up
thrust, or two of the three basic characteristics of the loop. They dont have the same "easy"
flow that plain arches do and particularly have significant up thrusts in the ridges near the
middle that arrange themselves on both sides of a spine or axis towards which the adjoining
ridges converge and appear to form tents.

Plain

Arch

Tented

Arch

Loops occur in about 60-70 % of fingerprint patterns encountered. One or more of the
ridges enters on either side of the impression, re-curves, touches or crosses the line running
from the delta to the core and terminates on or in the direction of the side where the ridge or
ridges entered. Each loop pattern has is one delta and one core and has a ridge count. Radial
loops are named after the radius, a bone in the forearm that joins the hand on the same side as
the thumb. The flow of the pattern in radial loops runs in the direction of the radius (toward
the thumb). Radial loops are not very common and most of the time radial loops will be
found on the index fingers. Ulnar loops are named after the ulna, a bone in the forearm. The
ulna is on the same side as the little finger and the flow of the pattern in a ulnar loop runs in
the direction of the ulna (toward the little finger).

Radial Loop

17

Ulnar Loop

18

Whorls are seen in about 25-35 % of fingerprint patterns encountered. In a whorl, some of
the ridges make a turn through at least one circuit. Any fingerprint pattern which contains 2
or more deltas will be a whorl pattern. There are four types of whorl patterns. Plain whorls
consist of one or more ridges which make or tend to make a complete circuit with two deltas,
between which an imaginary line is drawn and at least one re-curving ridge within the inner
pattern area is cut or touched. Central pocket loop whorls consist of at least one re-curving
ridge or an obstruction at right angles to the line of flow, with two deltas, between which
when an imaginary line is drawn, no re-curving ridge within the pattern area is cut or
touched. Central pocket loop whorl ridges make one complete circuit which may be spiral,
oval, circular or any variant of a circle. Double loop whorls consist of two separate and
distinct loop formations with two separate and distinct shoulders for each core, two deltas and
one or more ridges which make, a complete circuit. Between the two at least one re-curving
ridge within the inner pattern area is cut or touched when an imaginary line is drawn.
Accidental whorls consist of two different types of patterns with the exception of the plain
arch; have two or more deltas or a pattern which possess some of the requirements for two or
more different types or a pattern which conforms to none of the definitions.

Plain Whorl

Central Pocket Whorl

Double Loop Whorl

18

Accidental Whorl

19

3.3 History of Fingerprint Recognition:

3.3.1 William West:
The most famous case in the history of fingerprinting occurred in the late 19th century a man
was spotted in the incoming prisoner line at the U.S. Penitentiary in Leavenworth, Kansas by
a guard who 'knew' him and had just seen him already in the prison population. Upon
examination, the incoming prisoner claimed to be named Will West, while the (not escaped)
existing prisoner was named William West. According to their Bertillon measurements they
were essentially indistinguishable. As they were not twins, the Bertillon system came into
some question. However, their fingerprints were different, and fingerprint identification
received a significant boost in credibility.
Fingerprint imaging technology has been in existence for centuries. The use of fingerprints as
a unique human identifier dates back to second century B.C. China, where the identity of the
sender of an important document could be verified by his fingerprint impression in the wax
seal [3].The first modern use of fingerprints occurred in 1856 when Sir William Herschel, the
Chief magistrate of the Hooghly district in Jung poor, India, had a local businessman,
Rajyadhar Konia, impress his handprint on the back of a contract. Later, the right index and
middle fingers were printed next to the signature on all contracts made with the locals. The
purpose was to frighten the signer of repudiating the contract because the locals believed that
personal contact with the document made it more binding. As his fingerprint collection grew
Sir Herschel began to realize that fingerprints could prove or disprove identity.
The 19th century introduced systematic approaches to matching fingerprints to certain
individuals. One systematic approach, the Henry Classification System, based on patterns
such as loops and whorls, is still used today to organize fingerprint card files.
In the late 1960s NEC worked with the FBI and the Home Office in London, which had been
working on a system for New Scotland Yard from the late 1960s, to eventually develop a
minutia-based fingerprint identification system. It was initially installed in Tokyo in 1981 and
in San Francisco in 1983.
The first country to adopt a national computerized form of fingerprint imaging was Australia
in 1986, which implemented fingerprint imaging technology into its law enforcement system.

19

20

In 1996 after nearly a year of study, the National Institute of Standards and Technology has
been convinced that minutia is an acceptable way to store fingerprint biometric data on smart
cards. With the NIST acceptance of minutia it became inevitable this would set an industry
standard.
3.3.2 Minutia-Based Algorithm
Minutia-based algorithms extract information such as ridge ending, bifurcation, and short
ridge from a fingerprint image.

Short Ridge

Ridge Ending

Bifurcation

(Image source Wikipedia)

These features are then stored as mathematical templates. The identification or verification
process compares the template of the live image with a database of enrolled templates
(identification), or with a single enrolled template (authentication).

(Image source National Institute of Standards and Technology)

20

21

People with no or few minutia points (special skin conditions) cannot enroll or use the system
effectively. This is exemplified by the fingerprint immigration programs where finger
moistening peripherals are standard. Moreover, a low number of minutia points can be a
limiting factor for security of the algorithm. This can lead to false minutia points (areas of
obfuscation that appear due to low-quality enrollment, imaging, or fingerprint ridge detail). In
an application environment, enrollment without assistance may take several attempts due to
poor position or lack of pressure. While not quantified, user frustration will certainly have a
negative impact on technology acceptance.
3.3.3 Spectrum Analysis
Utilizing research from Nagoya Institute of Technology Graduate School in Japan, DDS has
developed an algorithm based on Spectrum Analysis. This technique captures cross sections
of a sliced fingerprint pattern and converts them to waves. Spectrum analysis uses the
spectral series of the waves as feature information, finding the maximum correlations in the
wave and verifies the identity of the fingerprint.

This algorithm of spectrum analysis works extremely well because this algorithm extracts
characteristics from the concavo-convex information of a fingerprint without being
influenced by the position of the characteristic points used in the conventional Minutia and
Pattern-matching method.
In the course of verification under the spectrum analysis algorithm, it is not necessary to store
the fingerprint image itself in the system which eliminates the possibility of exposure or
leakage of fingerprint images. In principle, it is impossible to regenerate original fingerprint
image from the extracted characteristics of images. This addresses issues raised by the IEEE
on fingerprint reconstruction of minutiae based systems.

21

22

The algorithm performs extremely well in controlled environments where positioning of the
finger in enrollment and verification are similar. However, with disparate fingerprint
positioning for enrollment and verifications results can be less than desired. This requirement
limits the application developer to more controlled ergonomic environments and may reduce
some commercial viability.
3.3.4 The Hybrid technology:
To accentuate the strengths of both the Spectrum Analysis and Minutiae-based algorithm and
limited the inherent weaknesses, the company has combined both algorithms and created a
Hybrid algorithm. The combined technology provides for rapid and accurate enrollment and
verification in difficult environments.
This Hybrid algorithm extends beyond combining scores from both techniques to form a
single decision; instead the technology utilizes a proprietary technique of Shading.
Shading analyzes the scores of both results and places and associates an importance value
with each. Using an algorithm based upon a database of past results, a final score is
calculated by using the individual results as a function of importance. With shading, if one
score is high and one is low, more decision weighting is placed upon the higher score. If both
scores are low then information from both are weighed more equally and the results are
combined together for a final decision. This ability allows for both algorithms to have low
scores and still be accurate.

22

23

The technology addresses both the strengths of both as well as the challenges. Spectrum
analysis can work well with poor image quality and difficult to read fingerprints and
minutiae-based algorithms can work very well with angles.

Hybrid
Example

23

Enrollment

24

FINGERPRINT RECOGNITION

24

25

4. FINGER PRINT RECOGNITION

4.1 PRE PROCESSING
This is an essential part of finger print recognition, in this step the image is made ready for
the actual matching, the input of this phase is original fingerprint image and the final output
of this step is minutiae of that image.
Our proposed algorithm is as follows.

4.1.1 Image enhancement:

The aim of image enhancement is to improve the interpretability or perception of information
in images for human viewers, or to provide `better' input for other automated image
processing techniques.
Image enhancement techniques can be divided into two broad categories:
1. Spatial domain methods, which operate directly on pixels, and
2. Frequency domain methods, which operate on the Fourier transform of an image.
Three types of degradations affect on quality of the fingerprint image, the ridges gets some
gaps: parallel ridges connected due to noise and natural effect to the finger like cuts, wrinkles
and injuries. The finger print enhancement is anticipated to improve the contrast between
ridges and valleys and reduce noises in the fingerprint images.
High quality finger print image is very important for fingerprint verification or identification
to work properly. In real life, the quality of the fingerprint image is affected by noise like
smudgy area created by over-inked area, breaks in ridges created by under-inked area,
changing the positional characteristics of fingerprint features due to skin resilient in nature,
dry skin leads to fragmented and low contrast ridges, wounds may cause ridge discontinuities
and sweat on fingerprints also leads to smudge marks and connects parallel ridges.

25

26

Following figures shows A. original finger print image and B. enhanced image.

4.1.2 Noise reduction

Noise is an unwanted perturbation to a wanted signal. Image noise is generally regarded as an
undesirable by-product of image capture. Noise reduction is the process of removing noise
from a picture (here it is the fingerprint image).We have checked and used different types of
filtering methods like median filter, global and adaptive thresholding to reduce the noise.

4.1.3 Binarization
In the pre-processing stage, the image is converted from grayscale to black and white. This is
done by calculating the average background intensity and subtracting this value from the
grayscale image.
Next grey scale threshold (basic global and adaptive thresholding) is calculated so pixels
above this value become black, and the ones below become white.

26

27

Figure: (a) Original Fingerprint (b) Binarized Fingerprint.

4.1.4 Thinning
Next the ridges must be thinned to a width of one-pixel. In this step two consecutive fast
parallel thinning algorithms are applied, in order to reduce to a single pixel the width of the
ridges in the binary image.
These operations are necessary to simplify the subsequent structural analysis of the image for
the extraction of the fingerprint minutiae. The thinning must be performed without modifying
the original ridge structure of the image.
During this process, the algorithms cannot miscalculate beginnings, endings and or
bifurcation of the ridges, neither ridge can be broken.

27

28

Figure shows the thinned image of the binarized image

4.2 SEGMENTATION:
In computer vision, segmentation refers to the process of partitioning a digital image into
multiple segments (sets of pixels, also known as super pixels). The goal of segmentation is to
simplify and/or change the representation of an image into something that is more meaningful
and easier to analyze.
Image segmentation is typically used to locate objects and boundaries (lines, curves, etc.) in
images. More precisely, image segmentation is the process of assigning a label to every pixel
in an image such that pixels with the same label share certain visual characteristics.
The result of image segmentation is a set of segments that collectively cover the entire image,
or a set of contours extracted from the image. Each of the pixels in a region is similar with
respect to some characteristic or computed property, such as color, intensity, or texture.
Adjacent regions are significantly different with respect to the same characteristic(s).
In this particular method, we first reading the particular image, and figure is shown by
appropriate command of mat lab. Then we moves for convolution.

28

29

Convolution computes the two-dimensional convolution of matrices A and B. If one of these

matrices describes a two-dimensional finite impulse response (FIR) filter, the other matrix is
filtered in two dimensions.
The size of C (convolution) in each dimension is equal to the sum of the corresponding
dimensions of the input matrices, minus one. That is, if the size of A is [ma, na] and the size
of B is [mb, nb], then the size of C is [ma+mb-1, na+nb-1].
Convolution for two dimensional matrices is given below.

C = conv2 (A, B)

In our finger print reorganization convolution is takes place between main image and ones
matrix (in this all rows and columns all denoted by 1) image. The resultant image is shown by
using figure.
In below shown figure, first image is original image and second image is segmented image.
In second image is portioned as black and white, i.e. the image is represented in white color.

29

30

4.3 Feature extraction

In pattern recognition and in image processing, feature extraction is a special form of
dimensionality reduction. This is a part of image processing, where a part of the whole image
is taken to recognize its specific pattern and then it is measured to classify the image on the
basis of that particular measurement.
Normally the second step in image analysis that seeks to measure the individual features
of the blobs or objects in the scene. Image features at various levels of complexity are
extracted from the image data. Typical examples of such features are Lines, edges and ridges.
Localized interest points such as corners, blobs or points.

4.3.1 Global extraction

Many object recognition systems use global features that describe an entire image. Most
shape and texture descriptive fall into this category; such features are attractive because they
produce very compact representations of images, where each image corresponds to a point in
a high dimensional feature space. Global features are sensitive to clutter and occlusion.

4.3.1.1 Mean and standard deviation of binary image:

In this algorithm we used very easy methods for better understanding; we commonly use
mean and standard deviation, which is familiar and implementing easily.
In firstly images and portioning into eight groups by adding noise, and forming as individual
images. Now the image reading is done, and it is converted into gray to binary. Gray is a
color of combination of black to white which has 0 to 255 pixels.
Binarization is a process of converting only in two colors; they are black and white, 1 for
white and 0 for black.

30

31

Figure: (a) Original Fingerprint (b) Binarized Fingerprint.

Applying the mean (average) and standard deviation of each image by different ways i.e. by
normal, increasing size, decreasing size, horizontal rotation, vertical rotation and rotation in
particular angle. Mean values are shown in table below
Image

Normal

Zoom

Horizontal

Vertical

Rotation

0.9207

0.9207

0.9207

0.9207

0.9207

0.9576

0.9576

0.9576

0.9576

0.9576

0.9997

0.9997

0.9997

0.9997

0.9997

0.8733

0.8733

0.8733

0.8733

0.8733

0.9700

0.9700

0.9700

0.9700

0.9700

0.9515

0.9515

0.9515

0.9515

0.9515

0.9245

0.9245

0.9245

0.9245

0.9245

0.9950

0.9950

0.9950

0.9950

0.9950

For standard deviation values are shown in table below.

31

32

Image

Normal

Zoom

Horizontal

Vertical

Rotation

0.1722

0.1722

0.1722

0.1722

0.1722

0.1665

0.1665

0.1665

0.1665

0.1665

0.2241

0.2241

0.2241

0.2241

0.2241

0.0132

0.0132

0.0132

0.0132

0.0132

0.1348

0.1348

0.1348

0.1348

0.1348

0.0794

0.0794

0.0794

0.0794

0.0794

0.0099

0.0099

0.0099

0.0099

0.0099

0.0066

0.0066

0.0066

0.0066

0.0066

For matching taking the average of every group values and checking each value in it.
This method is not sufficient for feature extraction of matching fingerprint; binary image
contains only 50% of white and 50% of black. The mean of that is may be 50 it and STD is
relatively very small, if some images having poor binarization so that will get insufficient
values. Due to this reason we are going to use direct image in feature extraction.

4.3.1.2 Mean and standard deviation of gray image

In this method, mean and standard deviation is calculated for gray images of respected
groups. Gray image is a combination of black to white colors ranging from 0 to 255.0 for
black and up to 255 is different combination of colors. (All values of each group are
averaging them for an individual value to represent a group value).

The details of mean values are shown in below table.

32

33

Image

Normal

Zoom

Horizontal

Vertical

Rotation

178.1069

178.1069

178.1069

178.1069

178.1069

177.3688

177.3688

177.3688

177.3688

177.3688

177.4977

177.4977

177.4977

177.4977

177.4977

198.3028

198.3028

198.3028

198.3028

198.3028

187.2216

187.2216

187.2216

187.2216

187.2216

197.4129

197.4129

197.4129

197.4129

197.4129

214.1794

214.1794

214.1794

214.1794

214.1794

226.2012

226.2012

226.2012

226.2012

226.2012

The details of standard deviation are as shown in below table.

33

Image

Normal

Zoom

Horizontal

Vertical

Rotation

33.1452

33.1452

33.1452

33.1452

33.1452

33.1313

33.1313

33.1313

33.1313

33.1313

33.8583

33.8583

33.8583

33.8583

33.8583

30.5665

30.5665

30.5665

30.5665

30.5665

24.4904

24.4904

24.4904

24.4904

24.4904

26.5039

26.5039

26.5039

26.5039

26.5039

19.5420

19.5420

19.5420

19.5420

19.5420

21.6679

21.6679

21.6679

21.6679

21.6679

34

After finding the values of each group result, matching is performed by comparing those
values to individual images.
As compared to above method it gives the better result.

4.3.2 Local extraction

Local feature is an image pattern which differs from its immediate neighborhood. It is
usually associated with a change of image property or several properties simultaneously. The
image properties considered are intensity color and texture.
In this process we are calculating by dividing whole image into some equal parts which is
referred as window. Window size will be 3, 5, 7 ..etc.
It is again divided into 2 types; they are level 1 and level 2.

4.3.2.1 Level 1
In this type of windowing sequences calculating only mean and STD is directly. Finding the
mean and variance is gives the value of that particular part of the image. So using it is not
effective of finding feature.
In this method, we compute the mean and variance for particular window size.

4.3.2.2 Level 2
In this type of widowing sequence along with the mean and STD, again calculating mean of
mean, mean of STD ,STD of mean , STD of STD. it is a level two process.
In this section we are using level 2 for better results. While performing mean of mean stands
for mean of the total image which is portioned into some small images, and to obtain a single
value. STD a measure of the variation between individuals on a variables, the variance is
used as a measure of how far a set of numbers are spread out from each other. It is one of
several descriptors of a probability distribution, describing how far the numbers lie from the
mean (expected value). In particular, the variance is one of the moments of a distribution. In
that context, it forms part of a systematic approach to distinguishing between probability
distributions. While other such approaches have been developed, those based on moments are
advantageous in terms of mathematical and computational simplicity.

34

35

4.3.3 Data base

It is of two types, they are mentioned as follows.

4.3.3.1. Average of images and finding features:

Averaging images and finding the values of that image. But this process using in finger print
is not sufficient, because in fingerprints orientations are different. It is mainly used in iris
recognition and face recognition which has less orientation (in this recognition systems first
averaging images and then finding its feature values).

4.3.3.2. Finding feature and then averaging values:

In this method first finding feature and then averaging values is nothing but first finding the
values of mean and variance for individual values and finding average of those values.
In this process we are using level 2 form for better results, first finding mean and variance
for particular window size ,and then again finding the mean of mean ,mean of STD,STD of
mean and STD of STD.

For level 1:
In this method mean &standard deviation of a fingerprint image are calculated for every
individual image of the group and then average the mean &standard deviation of each person
to form a one value for every group representation. Arrange them in matrix form. The matrix
is as follows.

35

36

Person

average of mean

average of standard deviation

101

142.5927

60.3239

102

135.6616

70.4338

103

140.8924

66.7694

104

150.3288

63.0280

105

179.7887

34.4733

106

134.5820

62.1796

107

112.7660

65.2008

108

155.1999

70.8154

The matrix loaded into the mat lab.

For image matching take the fingerprint image of mean & standard deviation are calculated.
Check the values in the above matrix. If that value is nearly equals to the any of the row of a
matrix. Then that fingerprint image is matching to that particular group.
This method is not suited for the detection of fingerprint. Because the possibilities of
matching is low so, level2 method is preferred.

For level2:
In this method finger print image is divided into 5*5 windows or 3*3 windows or 7*7
windows. Find the mean and standard deviation of the window. The values of mean is stored
in new image the values of standard deviation is stored in another new image. Find the mean
&standard deviation of mean image. Find the mean &standard deviation of standard
deviation image. So, we get 4 values.
The 4 values are mean of mean, mean of standard deviation, standard deviation of mean,
standard deviation of standard deviation. The values are arranged into the form of matrix.

36

37

The matrix for 3*3 windows

Person

Value1

Value2

Value3

Value4

101

182.7476

8.5417

33.5428

4.7375

102

180.2235

8.6952

30.6629

5.1416

103

194.8966

7.8224

29.5237

4.7500

104

205.6726

8.2255

27.1873

4.7734

105

213.1428

6.5175

26.2242

4.5586

106

213.1428

6.5175

26.2242

4.5586

107

217.3441

5.4815

24.1845

3.8095

108

185.9957

6.6812

32.1824

4.3552

The matrix for 5*5 windows

Person

Value1

Value2

Value3

Value4

101

181.0733

12.7290

34.9037

6.2025

102

178.6504

12.5450

32.3001

6.5500

103

193.2517

11.5380

31.9981

6.2999

104

204.0442

12.0221

30.3596

6.1085

105

181.2084

10.3635

36.5220

5.5687

106

211.4820

9.2420

30.4348

6.0274

107

215.6881

7.9292

29.1469

5.0887

108

184.3521

9.5137

34.5525

5.8304

The matrix for 7*7 windows

37

38

Person

Value1

Value2

Value3

Value4

101

179.4063

33.5428

8.5417

4.7375

102

180.2235

30.6629

8.6952

5.1416

103

194.8966

29.5237

8.224

4.7500

104

205.6726

27.1873

8.2255

4.7734

105

182.9563

34.6545

7.1993

4.3590

106

213.1428

26.2242

6.5175

4.5586

107

217.3441

24.1845

5.4815

3.8095

108

185.9957

32.1824

6.6812

4.3552

The above matrices are loaded into the mat lab for processing these matrices are saved in
individual way. And for matching process those values are loaded into mat lab and compared
with any finger print image .in this process we find the minimum distance to that entire
matrix and image values. In this process, if the value of image is near to any row we conclude
that it is from that group.
Here we are using different window sizes for every window size the probability of matching
is different. It is shown in below table.
Window sizes

Probability of

Probability of

matching correct
60%

matching wrong
40%

70%

30%

80%

20%

From the above table, window 7*7 is obtaining correct is greater than 3*3 and 5*5

38

39

5. CLASSIFICATION OF SCATTERING

5. CLASSIFICATION OF SCATTERING
39

40

CLUSTERING TECHNIQUE:
Cluster analysis, also called segmentation analysis or taxonomy analysis, creates groups, or
clusters, of data. Clusters are formed in such a way that objects in the same cluster are very
similar and objects indifferent clusters are very distinct. Measures of similarity depend on the
application.
'Cluster analysis' is a class of statistical techniques that can be applied to data that exhibit
natural groupings. Cluster analysis sorts through the raw data and groups them into
clusters. A cluster is a group of relatively homogeneous cases or observations. Objects in a
cluster are similar to each other. They are also dissimilar to objects outside the cluster,
particularly objects in other clusters.
The diagram below illustrates the results of a survey that studied drinkers perceptions of
spirits (alcohol). Each point represents the results from one respondent. The research
indicates there are four clusters in this market.

This method is described as follows.

40

41

Firstly image is read by using its appropriate command and it is converted into binary image.
1. Formulate the problem - select the variables to which you wish to apply the clustering
technique
2. Select a distance measure- various ways of computing distance:
o

Squared Euclidean distance - the square root of the sum of the squared
differences in value for each variable

Manhattan distance - the sum of the absolute differences in value for any
variable

Chebyshev distance - the maximum absolute difference in values for any

variable

Mahalanobis (or correlation) distance - this measure uses the correlation

coefficients between the observations and uses that as a measure to cluster
them. This is an important measure since it is unit invariant (can figuratively
compare apples to oranges)

3. Select a clustering procedure (see below)

4. Decide on the number of clusters
5. Map and interpret clusters - draw conclusions - illustrative techniques like perceptual
maps, icicle plots, and dendrograms are useful
6. Assess reliability and validity - various methods:

41

repeat analysis but use different distance measure

repeat analysis but use different clustering technique

split the data randomly into two halves and analyze each part separately

repeat analysis several times, deleting one variable each time

repeat analysis several times, using a different order each time

42

In our method we are finding distance from selected image to cluster by using distance
transform of binary image.
Syntax,
D = bwdist (BW)
[D, L] = bwdist (BW)
[D, L] = bwdist (BW, method)
Methods include is as shown below.
1. Chessboard: In 2-D, the chessboard distance between (x1, y1) and (x2,y2) is
Ch=max ([x1-x2], [y1-y2])
2. 'City block: In 2-D, the city block distance between (x1, y1) and (x2, y2) is
Ct=[x1-x2] + [y1-y2]
3.Euclidean': In 2-D, the Euclidean distance between (x1, y1) and (x2, y2) is
Ec=sqrt ((x1-x2) ^2+ (y1-y2) ^2)
This is the default method.

4.Quasi-euclidean': In 2-D, the quasi-Euclidean distance between (x1,y1) and (x2,y2) is

Qec=[x1-x2] + (sqrt (2)-1) [y1-y2], [x1-x2]> [y1-y2]
(Sqrt (2)-1)[X1-x2]+ [y1-y2], otherwise
42

43

6. CONCLUSION AND FUTURE SCOPE

43

44

6. CONCLUSION AND FUTURE SCOPE

Fingerprint Image enhancement is to make the image clearer for easy further operations.
Since the fingerprint images acquired from sensors or other media are not assured with
perfect quality, enhancement methods, for increasing the contrast between ridges and
furrows and for connecting the false broken points of ridges due to insufficient amount of
ink, are very useful to keep a higher accuracy to fingerprint recognition. Two methods are
adopted in the work: the first one is Histogram Equalization; the second one is Fourier
Transform.
1. Region of interest tells us which region of the fingerprint is very useful for various features
matching.
2. Improved thinning in the present work contributes to: The Image becomes perfectly
thinned to single pixel width. More number of bifurcations can be detected.
3. Also a program coding with MATLAB going through all the stages of the fingerprint
recognition is built. It is helpful to understand the procedures of fingerprint recognition. And
demonstrate the key issues of fingerprint recognition.
Overall, a set of reliable techniques have implemented for fingerprint recognition. These
techniques can then be used to facilitate the further study of the statistics of fingerprints. The
future scope of the work is to do Binarization of the image, to find the direction of the
bifurcation and ridges of the image, to extract features, to find actual minutiae .So that the
fingerprint recognition could be made better which can improve the future stages and the
final outcome.

44

45

7. REFFERENCES

45

46

7. REFFERENCES

1. Jain, L.C. et al. (Eds.). 1999. Intelligent Biometric Techniques in Fingerprint and
Face Recognition. Boca Raton, FL: CRC Press.
2. Lunenburg, Glenn (January 24, 2005). "Are one's fingerprints similar to those of his
or her parents in any discernable way?". Scientific American. Retrieved 28 August
2010.
3. Thornton, John (May 9, 2000). "Latent Fingerprints, Setting Standards In The
Comparison and Identification". 84th Annual Training Conference of the California
State Division of IAI.. Retrieved 30 August 2010.
4. Diaz, Raul (2007). "Biometrics: Security Vs Convenience". Security World
Magazine.. Retrieved 30 August 2010.
5. Meghdadi, Majid; Jalilzadeh, Saeed (29 October 2005). "Validity and Acceptability of
Results in Fingerprint Scanners". Proceedings of the 7th WSEAS International
Conference on Mathematical Methods and Computational Techniques In Electrical
Engineering. World Scientific and Engineering Academy and Society.. Retrieved 4
November 2010.
6. Setlak, Dale. "Advances in Biometric Fingerprint Technology are Driving Rapid
Adoption in Consumer Marketplace". Authentic. Retrieved 4 November 2010.
7. Mazumdar, Subhra; Dhulipala, Venkata (2008). "Biometric Security Using Finger
Print Recognition"(PDF). University of California, San Diego. p. 3.. Retrieved 30
August 2010.
8. Minutia vs. Pattern Based Fingerprint Templates. (2003). Retrieved December 13,
2005, from (archived from www.ibia.org on 2007-09-29)

46

47

47