14EM1D4702.pdf

(1)

BASED LIVENESS DETECTION TECHNIQUE

A THESIS REPORT

Submitted in partial fulfillment of the requirement to

JAWAHARLAL NEHRU TECHNOLOGICAL UNIVERSITY KAKINADA

For the Award of the Degree of

Master of Technology

in

Communication Systems

Submitted by

K. APARNA JYOTHI

(Regd. No: 14EM1D4702) Under the Esteemed Guidance of

Mr. K.VENKATESULU, M.Tech.,

Assistant Professor in ECE

DEPARTMENT OF ELECTRONICS & COMMUNICATION ENGINEERING

SWARNANDHRA INSTITUTE OF ENGINEERING AND TECHNOLOGY Accredited by NAAC with ‘A’ Grade

(Approved by A.I.C.T.E & Affiliated to JNTUK Kakinada) Seetharampuram, Narsapur-534280,

West Godavari (Dt.), A.P. 2015-2016

(2)

ENGINEERING

SWARNANDHRA INSTITUTE OF ENGINEERING AND TECHNOLOGY

Accredited by NAAC with ‘A’ Grade

(Approved by A.I.C.T.E & Affiliated to JNTU Kakinada) Seetharampuram, Narsapur-534280,

West Godavari (Dt.), A.P.

CERTIFICATE

This is to certify that this dissertation work entitled “

Fake Biometric

Detection Using Software based Liveness Detection Technique

” being

submitted by K. APARNA JYOTHI (14EM1D4702) in partial fulfillment of the requirements for the award of Master of Technology with COMMUNICATION SYSTEMS as specialization is a record of bonafide work carried out by him, during the academic year 2015 – 2016.

Project Guide Head of the Department Mr. K.Venkatesulu, M.Tech., Mr.V.Srinivas, M.Tech., (Ph.D)

Assistant Professor, ECE Associate Professor, ECE

(3)

i

ACKNOWLEDGEMENT

The satisfaction that accompanies the successful completion of every task during my dissertation would be incomplete without the mention of the people who made it possible. I consider it my privilege to express my gratitude and respect to all who guided, inspired and helped me in completion of my project work.

I extend my heartfelt gratitude to the Almighty for giving me strength in proceeding with this project titled “Fake Biometric Detection Using Software Based Liveness Detection Technique”.

I owe my special thanks to Dr. S. Ramesh Babu, M.Tech.,Ph.D., Secretary & Correspondent,Swarnandhra College of Engineering and Technology, Seetharampuram for providing necessary arrangements to carry out this project.

I express my heartfelt thanks to Dr.T.Madhu, M.Tech., Ph.D.FIE.,MISTE, Principal, Swarnandhra College of Engineering & Technology, Seetharampuram for giving me this opportunity for the successful completion of my degree.

I would like to express my sincere thanks to Mr.V.Srinivas,M.Tech.(Ph.D)., Associate Professor & Head of the Department of ECE for his valuable suggestions at the time of need.

I record with pleasure, of my gratitude to my Project guide Mr.K.Venkatesulu, M.Tech., Assistant Professor, ECE for his simulating guidance and valuable suggestions throughout the project.

I am very much grateful to our M.Tech Coordinator Mr.S.Ravichand, M.Tech.(Ph.D)., Associate professor for giving the encouragement that helped me to complete the project successfully.

I would like to express my profound sense of gratitude to all the coordinators and faculty members for their cooperation and encouragement throughout my course. Above all, I thank my parents. I feel deep sense of gratitude for my family who formed part of my vision. Finally I thank one and all that have contributed directly or indirectly to this thesis.

K.Aparna jyothi

(14EM1D4702)

(4)

ii

ABSTRACT

Biometrics-based authentication systems offer obvious usability advantages over traditional password and token-based authentication schemes. However, biometrics raises several privacy concerns. A biometric is permanently associated with a user and cannot be changed. Hence, if a biometric identifier is compromised, it is lost forever and possibly for every application where the biometric is used. Moreover, if the same biometric is used in multiple applications, a user can potentially be tracked from one application to the next by cross-matching biometric databases. In this PROJECT, we demonstrate several methods to generate multiple cancelable identifiers from fingerprint images to overcome these problems. Designing of a novel software-based fake detection method that can be used in multiple biometric systems to detect different types of fraudulent access attempts. We outlined several advances that originated both from the cryptographic and biometric community to address this problem. In particular, we outlined the advantages of cancelable biometrics over other approaches and presented a case study of different techniques.This project can be enhanced by reducing the image using DTCWT. This modification can decrease the image size and execution can be reduced by enhancing the image clarity. This enhancement can be shown using PSNR values.

The experimental results, obtained on publicly available data sets of fingerprint, iris, and 2D face, show that the proposed method is highly competitive compared with other state-of-the-art approaches and that the analysis of the general image quality of real biometric samples reveals highly valuable information that may be very efficiently used to discriminate them from fake traits.

(5)

iii

LIST OF CONTENTS

Name Of The Content Page No.

ACKNOWLEDGEMENT i

ABSTRACT ii

LIST OF CONTENTS iii

LIST OF FIGURES v

LIST OF TABLES vi

CHAPTER 1: INTRODUCTION 1

1.1 INTRODUCTION 2

1.2 AIM AND OBJECTIVES 3

1.3 LITERATURE SURVEY 4

1.4 ORGANISATION REPORT 5

CHAPTER 2: EXISTING SYSTEM 6

2.1 EXISTING SYSTEM 7

2.2 DRAWBACKS OF EXISTING SYSTEM 8

2.2.1 Duplication with co-operation 8

2.2.2 Duplication without Co-operation 9

CHAPTER 3: PROPOSED SYSTEM 12

3.1 BIOMETRIC AUTHENTICATION 12

3.2 FINGER PRINT RECOGNIZATION 15

3.2.1 Different patterns 15

3.2.2 Minutia features 16

3.3 IRIS 18

3.3.1 Iris image enhancement 18

3.3.2 Iris image Binarization 19

3.3.3 Iris Image Segmentation 19

3.4 MINUTIA Extraction 19

3.4.1 Iris Ridge Thinning 19

3.4.2 Enhanced Thinning 20

3.4.3Enhanced thinning algorithm 21

3.5 FACE RECOGNITION 21

(6)

iv

4.1 LIVENESS ASSESSMENT IN AUTHENTICATION SYSTEM 25

4.2 SOFTWARE-BASED TECHNIQUES 25

4.3 IMAGE QUALITY ASSESSMENT FOR LIVENESS

DETECTION 26

4.4 FULL-REFERENCE IQ MEASURES 28

4.4.1 Different types FR-IQ Measurements 30

4.5 NO-REFERENCE IQ MEASURES 33

4.6 DUAL TREE COMPLEX WAVELETS TRANSFORM 36

CHAPTER 5: SOFTWARE ASPECTS 39

5.1 INTRODUCTION TO MATLAB 40

CHAPTER 6: RESULTS 47

6.1 TOP MODULE 48

6.2 INPUT PROCESS WINDOW 49

6.3 IDENTIFICATION PROCESS 50

6.4 IQA PARAMETERS 51

6.5 FINAL RESULT 52

6.6

Advantages

53

6.7 Applications 53

CHAPTER 7: CONCLUSION & FUTURE SCOPE 54

7.1 CONCLUSION 55 7.2 FUTURE SCOPE 55 REFERENCES 56 BIBLIOGRAPHY 59 APPENDIX 61

(7)

v

LIST OF FIGURES

Figure No. Name Of The Figure Page No.

2.1 software-based liveness detection 7

2.2 A Stamp type dummy finger print 9

2.3 A wafer-thin silicon dummy fingerprint 9

3.1 Block diagram of Biometric system 13

3.2(a) The Arch Pattern 15

3.2(b) The Loop Pattern 16

3.2(c) The Whirl pattern 16

3.3(a) Ridge Ending 17

3.3(b) Bifurcation 17

3.3(c) Short Ridge 17

3.4 Iris Real and Fake Images 19

4.1 Block diagram for the proposed system 25

4.2 Classification Of Image Quality Measures 28

4.3 Block diagram for a 3-level DTCWT 36

5.1 Example image (f) 41

5.2 Example picture (rice.png) 45

6.1 Top Module 49

6.2 Input Process Window 50

6.3 Identification Process 51

6.4 Measured IQ Parameters 52

6.5 Final Result 53

(8)

vi

LIST OF TABLES

Table No. Name of the Table Page No.

4.1 Different parameters measured for an image 35

(9)

SWARNANDHRA INSTITUTE OF ENGINEERING & TECHNOLOGY Page 1

CHAPTER 1 INTRODUCTION

(10)

CHAPTER 1 INTRODUCTION

1.1 INTRODUCTION

Biometrics refers to metrics related to human characteristics. Biometrics authentication (or realistic authentication) is used in computer science as a form of identification and access control. It is also used to identify individuals in groups that are under surveillance. Biometric identifiers are the distinctive, measurable characteristics used to label and describe individuals. Biometric identifiers are often categorized as physiological versus behavioural characteristics. Physiological characteristics are related to the shape of the body. Examples include, but are not limited to fingerprint, palm veins, face recognition, DNA, palm print, hand geometry, iris recognition, retina and odour/scent.

Behavioural characteristics are related to the pattern of behaviour of a person, including but not limited to typing rhythm, gait, and voice. Some researchers have said that the term behaviour metrics is used to describe the latter class of biometrics. More traditional means of access control include token-based identification systems, such as a driver's license or passport, and knowledge-based identification systems, such as a password or personal identification number.

Since biometric identifiers are unique to individuals, so they provide high reliability in verifying the identity than token and knowledge-based methods; however, the collection of biometric identifiers raises privacy concerns about the ultimate use of this information. Securing information and ensuring the privacy of personal identities is a growing concern in today‘s society. Traditional authentication schemes primarily utilize tokens or depend on some secret knowledge possessed by the user for verifying his or her identity.

Biometrics-based authentication schemes using fingerprints, face recognition, etc., overcome these limitations while offering usability advantages and are therefore rapidly extending traditional authentication schemes. However, despite its obvious advantages, the use of biometrics raises several security and privacy concerns as outlined below:

(11)

Biometrics is authentic but not secret: Unlike passwords and cryptographic keys that are known only to the user, biometrics such as voice, face, signature, and even fingerprints can be easily recorded and potentially misused without the user‘s consent. There have been several instances where artificial fingerprints have been used to circumvent biometric security systems. Face and voice biometrics are similarly vulnerable to being captured without the user‘s explicit knowledge. In contrast, tokens and knowledge have to be willingly shared by the user to be compromised.

Biometrics cannot be revoked or cancelled. Passwords, PINs, etc., can be reset if compromised. Tokens such as credit cards and badges can be replaced if stolen. However, biometrics is permanently associated with the user and cannot be revoked or replaced if compromised. While a user can successively enrol different fingerprints, there is still a limited choice of fingers to choose from. This choice does not exist for other biometric modalities.

If once the Biometric is lost, it is compromised forever, why because Biometrics provides usability advantages that it obviates the need to remember and manage multiple passwords and identities. However, this also means that if a biometric is compromised in one application, essentially all applications where the particular biometric is used are compromised. Cross-matching can be used to track individuals without their permission. Since the same biometric might be used for various applications and locations, the user can potentially be tracked if organizations share their respective biometric databases. So in order to provide security from this type of hackers, the traditional authentication schemes are used; by using this scheme the user can maintain different identities/passwords to prevent this hacking. The fact that a biometric remains the same presents a privacy concern. 1.2 AIM & OBJECTIVES

:

We demonstrate several methods to generate multiple cancellable identifiers from fingerprint images to overcome the hacking techniques. Designing of a novel software-based fake detection method that can be used in multiple biometric systems to detect different types of fraudulent access attempts. We outlined several advances that originated both from the cryptographic and biometric community to address this problem. In particular, here we mentioned the advantages of cancellable biometrics

(12)

over other approaches and presented a case study of different techniques. This project can be enhanced by reducing the image using DTCWT (Dual Tree Complex Wavelet Transform).

The objective of this project is modification can decrease the image size and execution can be reduced by enhancing the image clarity.

1.3 LITERATURE SURVEY:

After Rathaetal. formally defined the problem of cancellable biometrics (also called revocable biometrics), many alternate solutions have emerged from both the biometric and cryptographic community. We loosely divide the prior work into the following categories

 Biometric salting: This is similar to password salting in conventional crypto-systems. In this approach, before hashing the password P of the user, it is concatenated with a pseudorandom string S and the resulting hash is stored in the database. The addition of the random sequence increases the entropy and, therefore, the security of the password. Biometric salting is based on the same principle. In some instances, the new representation is quantized to derive robust binary cryptographic keys. However, the quantization is practical only because of the additional entropy introduced through the salt.In this category the defining feature is the addition of user-specific random information to increase the entropy of the biometric template. 



 Biometric key generation: In this approach, a key is derived directly from the biometric signal. The advantage is that there is no need for user-specific keys or tokens as required by ―biometric salting‖ methods and that it is therefore scalable. A key, parameterized by the biometric B, is stored instead of the actual biometric itself.The major problem with this approach is achieving error tolerance in the key. The defining feature of this category is the attempt to derive robust binary representations (keys) from noisy biometric data without the use of additional information. 



 Fuzzy schemes: Another approach for constructing cancellable templates involves the use of public auxiliary information P (also called helper data, shielding functions, or fuzzy extractors), which is combined with biometric information to reduce the intra user variation. The following is the defining feature of this category: The schemes define a metric (e.g., Hamming,

(13)

Euclidian, set distance, etc.) on noisy biometric data B and B0. Among the different threats analyzed, the so-called direct or spoofing attacks have motivated the biometric community to study the vulnerabilities against this type of fraudulent actions in modalities such as the iris, the fingerprint, the face, the signature, or even the gait and multimodal approaches.

1.4 ORGANISATION REPORT:

This thesis consists of total SEVEN chapters that include introduction and conclusions. Chapter 2 describes about existing system. in chapter 3 conceptual analysis of proposed system.chapter 4 having implementation of project. chapter 5 describes software aspects. Chapter 6 result analysis. in chapter 7 advantages and applications are discussed.

(14)

CHAPTER-2

EXISTING SYSTEM

(15)

CHAPTER 2 EXISTING SYSTEM

2.1 EXISTING SYSTEM:

In recent years, the increasing interest in the evaluation of biometric systems security has led to the creation of numerous and very diverse initiatives focused on Biometric systems. Among the different threats analyzed, the so-called direct or

spoofing attacks have motivated the biometric community to study the vulnerabilities

against this type of fraudulent actions in modalities such as the iris, the fingerprint, the face, the signature, or even the gait and multimodal approaches. In these attacks, the intruder uses some type of synthetically produced artefact (e.g., gummy finger, printed iris image or face mask), or tries to mimic the behaviour of the genuine user (e.g., gait, signature), to fraudulently access the biometric system. As these types of attacks are performed in the analog domain and the interaction with the device is done following the regular protocol, the usual digital protection mechanisms (e.g., encryption, digital signature or watermarking) are not effective.

Besides other anti-spoofing approaches such as the use of multi biometrics or challenge-response methods, special attention has been paid by the users to the

liveness detection techniques, which use different physiological properties to

distinguish between real and fake traits. Biometrics-based authentication systems offer obvious usability advantages over traditional password and token-based authentication schemes. However, biometrics raises several privacy concerns. A biometric is permanently associated with a user and cannot be changed.

(16)

Liveness detection methods are usually classified into one of two groups shown in figure 2.1.

(i) Hardware-based techniques, which add some specific device to the sensor in

order to detect particular properties of a living trait (e.g., fingerprint sweat, blood pressure, or specific reflection properties of the eye).

(ii) Software-based techniques, in this case the fake trait is detected once the

sample has been acquired with a standard sensor (i.e., features used to distinguish between real and fake traits are extracted from the biometric sample, and not from the trait itself).

The two types of methods present certain advantages and drawbacks over the other and, in general, a combination of both would be the most desirable protection approach to increase the security of biometric systems.

2.2 DRAWBACKS OF EXISTING SYSTEM:

In the existing system we are using hardware techniques for implementing biometrics. In Hardware-based techniques, which add some specific device to the sensor in order to detect particular properties of a living trait (e.g., fingerprint sweat, blood pressure, or specific reflection properties of the eye), as a coarse comparison, hardware-based schemes usually present a higher fake detection rate. Furthermore, as they operate directly on the acquired sample (and not on the biometric trait itself), and the drawbacks in existing system are

 More costlier,  

 More complexity,  

 High computational delay,  

 Harmful to the human. 

And also in hardware Biometric systems one can easily misuse the system by creating fake or dummy fingerprints and dummy face masks etc. for example creation of dummy finger prints are explained below:

Dummy fingerprints can be created in two ways (i) Duplication without Co-operation (ii) Duplication with Co-operation.

2.2.1 Duplication without Co-operation:

For duplication of a fingerprint without co-operation of the original person it is necessary to obtain a print of the finger from a glass or another surface. One of the best ways to obtain such a print could be the fingerprint scanner itself. If the scanner is cleaned before a person will be using it, an almost perfect print is left on the scanner surface since people tend to press their finger (which is the verification finger!) firmly on the scanner. Some

(17)

more expertise is required to create a dummy from such a print, but every dental technician has the skills and equipment to create one. This is an accurate description of how to create a dummy of the fingerprint. A picture of a stamp that is created using this method can be found in Figure 2.

Fig 2.2 A stamp type dummy finger print

2.2.2 Duplication with Co-operation:

Duplication of a fingerprint with co-operation of its owner is an easiest method since it is possible to compare the dummy with the original fingerprint in all aspects and adapt it accordingly. First, a plaster cast of the finger is created. This cast is then filled with silicon rubber to create a wafer-thin silicon dummy. This dummy can be glued to anyone's finger without it being noticeable to the eye. So one can easily create the dummy finger and can get the information without the intervention of original user. It follows that creation of this type of dummy is possible with very limited means within a few hours.

Fig 2.3: A wafer-thin silicon dummy fingerprint

To overcome this drawbacks, we demonstrate several methods to generate multiple cancelable identifiers from fingerprint images, that is Software Based techniques are used, in this case the fake trait is detected once the sample has been

(18)

acquired with a standard sensor (i.e., features used to distinguish between real and fake traits are extracted from the biometric sample, and not from the trait itself). In this several advances that originated both from the cryptographic and biometric community to address this problem.

As a comparison, hardware-based schemes usually present a higher fake detection rate, while software-based techniques are in general less expensive (as no extra device is needed), and less intrusive since their implementation is transparent to the user. Furthermore, as they operate directly on the acquired sample (and not on the biometric trait itself), software-based techniques may be embedded in the feature extractor module which makes them potentially capable of detecting other types of illegal break-in attempts not necessarily classified as spoofing attacks. For instance, software-based methods can protect the system against the injection of reconstructed or synthetic samples into the communication channel between the sensor and the feature extractor.

(19)

CHAPTER 3

(20)

CHAPTER 3 PROPOSED SYSTEM

The Proposed project can be enhanced from the existing technique by reducing the image using DTCWT (Dual Tree Complex Wavelet Transform). This modification can decrease the image size and execution can be reduced by enhancing the image clarity. This enhancement can be shown using PSNR (Peak Signal to Noise Ratio) values and provides high security to biometrics from the fake users. 3.1 BIOMETRIC AUTHENTICATION:

Many different aspects of human physiology, chemistry or behaviour can be used for biometric authentication. The selection of a particular biometric for use in a specific application involves a weighting of several factors. Jain et al. (1999) identified seven different factors which are used when assessing the suitability of any trait for use in biometric authentication. Universality means that every person using a system should possess the trait. Uniqueness means the trait should be sufficiently different for individuals in the relevant population such that they can be distinguished from one another. Permanence relates to the manner in which a trait varies over time. More specifically, a trait with 'good' permanence will be reasonably invariant over time with respect to the specific matching algorithm. Measurability (collectability) is nothing but ease of acquisition or measurement of the trait.

In addition, acquired data should be in a form that permits subsequent processing and extraction of the relevant feature sets. Performance relates to the accuracy, speed, and robustness of technology used (see performance section for more details). Acceptability relates to how well individuals in the relevant population accept the technology such that they are willing to have their biometric trait captured and assessed. Circumvention relates to the ease with which a trait might be imitated using an artifact or substitute.

The block diagram illustrates the two basic modes of a biometric system First, in verification (or authentication) mode the system performs a one-to-one comparison of a captured biometric with a specific template stored in a biometric database in order to verify the individual is the person they claim to be. Three steps are involved in the verification of a person.

(21)

indicate which template should be used for comparison 'Positive recognition' is a common use of the verification mode, "where the aim is to prevent multiple people from using same identity.

Fig 3.1 Block diagram of Biometric system

 In the first step, reference models for all the users are generated and stored

in the model database.  

 In the second step, some samples are matched with reference models to generate the genuine and impostor scores and calculate the threshold.  

 Third step is the testing step. 

Second is the identification mode, in identification mode the system performs a one-to-many comparison against a biometric database whether the given image is original or not (fake image). The system will succeed in identifying the individual if the comparison of the biometric sample to a template in the database falls within a previously set threshold. Identification mode can be used either for 'positive recognition' (so that the user does not have to provide any information about the template to be used) or for 'negative recognition' of the person "where the system establishes whether the person is who she (implicitly or explicitly) denies to be".

The latter function can only be achieved through biometrics since other methods of personal recognition such as passwords, PINs or keys are ineffective.

(22)

The first time an individual uses a biometric system is called enrollmen1t. During the enrolment, biometric information from an individual is captured and stored. In subsequent uses, biometric information is detected and compared with the information stored at the time of enrolment. Note that it is crucial that storage and retrieval of such systems themselves be secure if the biometric system is to be robust.

In the block diagram of Biometric system, the first block (sensor) is the interface between the real world and the system; it has to acquire all the necessary data. It is an image acquisition system, but it can change according to the characteristics desired.

The second block performs all the necessary pre-processing: it has to removing background noise from the sensor, to enhance the input, to use some kind of normalization, etc.

In the third block necessary features are extracted. This step is an important step as the correct features need to be extracted in the optimal way. A vector of numbers or an image with particular properties is used to create a template. A template is a synthesis of the relevant characteristics extracted from the source. Elements of the biometric measurement that are not used in the comparison algorithm are discarded in the template to reduce the file size and to protect the identity of the enrollee during the enrollment phase, the template is simply stored somewhere (on a card or within a database or both). During the matching phase, the obtained template is passed to a matcher that compares it with other existing templates, estimating the distance between them using any algorithm (e.g. Hamming distance).

The matching program will analyze the template with the input. This will then be output for any specified use or purpose (e.g. entrance in a restricted area) Selection of biometrics in any practical application depending upon the characteristic measurements and user requirements. We should consider Performance, Acceptability, Circumvention, Robustness, Population coverage, Size, Identity theft deterrence in selecting a particular biometric. Selection of biometric based on user requirement considers Sensor availability, Device availability, Computational time and reliability, Cost, Sensor area and power consumption.

(23)

3.2 FINGERPRINT RECOGNITION:

Fingerprint recognition or fingerprint authentication refers to the automated method of verifying a match between two human fingerprints. Generally most of biometrics which are manufactured at the early years based on finger recognization. Fingerprints are one of many forms of biometrics used to identify individuals and verify their identity. This article touches on two major classes of algorithms (minutia and pattern) and four sensor designs (optical, ultrasonic, passive capacitance, and active capacitance).

The analysis of fingerprints for matching purposes generally requires the comparison of several features of the print pattern. These include patterns, which are aggregate characteristics of ridges, and minutia points, which are unique features found within the patterns. But nowadays there are many ways to create a dummy fingerprints without permission of original one. So, in order to overcome all these cheating techniques we are using this liveness security system in biometrics. It is also necessary to know the structure and properties of human skin in order to successfully employ some of the imaging technologies.

3.2.1 Different patterns:

The three basic patterns of fingerprint ridges are the arch, loop, and whorl:

 Arch: The ridges enter from one side of the finger, rise in the center forming an arc, and then exit the other side of the finger. 



 Loop: The ridges enter from one side of a finger, form a curve, and then exit on that same side. 



 Whorl: Ridges form circularly around a central point on the finger. 

Scientists have found that family members often share the same general fingerprint patterns, leading to the belief that these patterns are inherited. The basic structure of three different fingerprint patters are showed below

(24)

3.2.2 Minutia features:

In the fingerprint recognization the second class of algorithm is minutia algorithm. As a global feature, orientation field describes one of the basic structures of a fingerprint. When the fingerprint is complemented with the minutiae, we can get more information. Thus, a better performance can be obtained by fusing the results of orientation field matching with conventional minutiae-based matching. Some studies showed that incorporating local (minutiae) and global (orientation field) feature can largely improve the performance.

However, as stated above, in many practical fingerprint recognition systems, the original images and orientation field images are not saved, and we cannot compute the orientation field directly.

In some other systems, additional orientation features cannot be saved into the existing database easily, and we have to compute the orientation field by only using the information of minutiae template. An interpolation algorithm is proposed to estimate the orientation field from minutiae template (they used it to predict the class of the fingerprint but not for fingerprint matching), in which the orientation of a given point was computed from its neighboring minutiae. To consider the global information, we will use the orientation model to reconstruct the orientation field from minutiae.

(25)

apply the model-based method on these mixed minutiae (including the real and virtual minutiae).

After that, the reconstructed orientation field is used into the matching stage by combining with conventional minutiae-based matching. In this way we have to create an oriented field fingerprint using minutia extraction techniques.

There are many different minutia extraction fields for different fingerprint ridges; here we use mainly three different types of minutia extraction techniques. Our present proposal is liveness detection method which uses this technique whether person who uses the system is original or not. Generally a person can cheat others but a machine not do that but nowadays the cheaters can try to cheat system also by preparing the dummy fingerprints. To protect the world from these unwanted users we are using the fingerprint recognization with minutia extraction fields.

When the fingerprint is complemented with the minutiae, we can get more information and better performance can be obtained by fusing the results in biometric systems. The three main minutia extraction fields used in fingerprint ridges are given below;

The major minatue features of fingerprint ridges are: Ridge ending and bifurcation &short ridge.

The ridge ending is the point at which a ridge terminates. Bifurcations are points at which a single ridge splits into two ridges. Short ridges (or dots) are ridges which are significantly shorter than the average ridge length on the fingerprint.

Minutiae and patterns are very important in the analysis of fingerprints since no two fingers have been shown to be identical.

(26)

fingerprint ridge detail come from the same finger. There exist multiple algorithms that do fingerprint matching in many different ways. Some methods involve matching minutiae points between the two images, while others look for similarities in the bigger structure of the fingerprint. In this project we propose a method for fingerprint matching based on minutiae matching.

However, unlike conventional minutiae matching algorithms our algorithm also takes into account region and line structures that exist between minutiae pairs. This allows for more structural information of the fingerprint to be accounted for thus resulting in stronger certainty of matching minutiae. Also, since most of the region analysis is pre-processed it does not make the algorithm slower. The Evidence from the testing of the pre-processed images gives stronger assurance that using such data could lead to faster and stronger matches.

3.3 IRIS:

The processing capabilities of the IRIS vision systems specifically using the IRIS v. The first stage of the architecture embeds sensors, parallel processing analog and mixed-signal circuitry, control circuitry and memory. This front-stage is implemented through dedicated bio-inspired chips. The second stage of the IRIS vision system architecture is a digital microprocessor. The combination of parallel preprocessing and serial post-processing makes the IRIS systems very efficient particularly the IRIS systems are capable to close the sensor-processing-actuation loop at a high speed. In this demo, the IRIS v is used to recognize data matrix codes at more than 200 codes/sec rate.

3.3.1 Iris image enhancement:

Iris Image enhancement is used to improve the image clarity for the easy of further operations. Since the iris 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 iris recognition. Two methods are used for the iris image enhancement, the first one is Histogram Equalization; the second one is Fourier Transform.

Histogram equalization: Histogram equalization is nothing but to separate the dynamic range and also provide the equal pixels in all the gray levels.

(27)

FT the original signal is represented in the form of infinite frequencies of varying amplitude. All the Images are periodic signals so FT are used to enhance the images.

3.3.2 Iris Image Binarization:

A locally adaptive binarization method is performed to binarize the iris image. Such a named method comes from the mechanism of transforming a pixel value to 1 if the value is larger than the mean intensity value of the current block (16x16) to which the pixel belongs.

3.3.3 Iris Image Segmentation:

In general, only a Region of Interest (ROI) is useful to be recognized for each iris image. The image area without effective ridges and furrows is first discarded since it only holds background information. Then the bound of the remaining effective area is sketched out since the minutia in the bounded region is confusing with those spurious minutia‘s that are generated when the ridges are out of the sensor. To extract the ROI, a two-step method is used. The first step is block direction estimation and direction variety check, while the second is intrigued from some Morphological methods. Two Morphological operations called OPEN and CLOSE are adopted. The OPEN operation can expand images and remove peaks introduced by background noise. The CLOSE operation can shrink images and eliminate small cavities.

Fig 3.4 Iris Real and Fake Images

3.4 MINUTIA EXTRACTION:

Minutiae Extraction steps are explained below 3.4.1 Iris Ridge Thinning:

(28)

just a single pixel width [5, 7]. The requirements of a good thinning algorithm with respect to an iris are

 The thinned iris image obtained should be of single pixel width with no discontinuities. 



 Each ridge should be thinned to its centre pixel.  

 Noise and singular pixels should be eliminated.  

 No further removal of pixels should be possible after completion of thinning process. 

Use an iterative, parallel thinning algorithm, in each scan of the full iris image, the algorithm marks down redundant pixels in each small image window (3x3). And finally removes all those marked pixels after several scans. But it is tested that such an iterative, parallel thinning algorithm has bad efficiency although it can get an ideal thinned ridge map after enough scans. Uses a one-in-all method to extract thinned ridges from gray-level iris images directly. Their method traces along the ridges having maximum gray intensity value. However, binarization is implicitly enforced since only pixels with maximum gray intensity value are remained.

The advancement of each trace step still has large computation complexity although it does not require the movement of pixel by pixel as in other thinning algorithms. Thus the third method is bid out which uses the built-in Morphological thinning function in MATLAB to do the thinning and after that an enhanced thinning algorithm is applied to obtain an accurately thinned image.

3.4.2 Enhanced Thinning:

Ridge Thinning is to eliminate the redundant pixels of ridges till the ridges are just one pixel wide. Ideally, the width of the skeleton should be strictly one pixel. However, this is not always true. There are still some locations, where the skeleton has a two-pixel width at some erroneous pixel locations. An erroneous pixel is defined as the one with more than two 4-connected neighbours. These erroneous pixels exist in the fork regions where bifurcations should be detected, but they have CN = 2 instead of CN>2. The existence of erroneous pixels may

 destroy the integrity of spurious bridges and spurs,  

 exchange the type of minutiae points, and  

(29)

Therefore, before minutiae extraction, there is a need to develop a validation algorithm to eliminate the erroneous pixels while preserving the skeleton connectivity at the fork regions. For this purpose an enhanced thinning algorithm is bid out.

3.4.3 Enhanced thinning algorithm:

Step 1: Scanning the skeleton of iris image row by row from top-left to bottom-right. Check if the pixel is 1.

Step 2: Count its four connected neighbours. Step 3: If the sum is 2+,then is erroneous pixel Step 4: Remove the erroneous pixel.

Step 5: Repeat steps 1 – 4 until whole of the image is scanned and the erroneous pixels are removed.

3.5 FACE RECOGNITION:

Multibiometrics refers to the use of a combination of two or more sensor modalities in a single identification system. The reason for combining different sensor modalities is to improve the recognition accuracy. A multisensory biometric system involving visual and thermal face images is present in this project. Face recognition is one of the most important applications of image analysis, its prime applications being recognition of individuals for the purpose of security. It is one of the most no obtrusive biometric techniques.

However, the conclusion was that, though the recognition performance of Even though faces recognition technology has moved from linear subspace methods. Eigen and Fisher faces, to nonlinear methods such as kernel principle component analysis (KPCA) and kernel Fischer discriminate analysis (KFDA), many of the problems are yet to be addressed. Also, the nature of research studies had been more on visible imagery with less attention on its thermal counterpart. Previous studies have shown that infrared imagery offers a promising alternative to visible imagery for handling variations in face appearance due to illumination changes.

It is observed that face recognition on thermal images degrades more sharply than with visible images when probe and gallery images are chosen from different sessions. Results in indicated better performance for visible imagery indoors under controlled lighting conditions, but thermal imagery-based face

(30)

recognition was better than its visible counterpart outdoors. Also, the recognition results of thermal imagery for both indoors and outdoors was found to be similar, thus proving that illumination had very little effect on thermal imagery.

Thermal imagery degraded the fusion of both visible and thermal modalities yielded better overall performance. Feature-based face recognition techniques have demonstrated the capability of invariance to facial variations caused by illumination and have achieved high accuracy rates. To make the recognition process illumination invariant, phase congruency feature maps are used instead of intensity values as the input to the face recognition system.

The feature selection process presented in this PROJECT is derived from the concept of modular spaces. Recognition techniques based on local regions have achieved high accuracy rates. Though the face images are affected due to variations such as non uniform illumination, expressions and partial occlusions, facial variations are confined mostly to local regions. Modularizing the images would help to localize these variations, provided the modules created are sufficiently small. But in this process, a large amount of dependencies among various neighboring pixels might be ignored. This can be countered by making the modules larger, but this would result in an improper localization of the facial variations.

In order to deal with this problem, a module creation strategy has been implemented in this PROJECT which considers additional pixel dependencies across various sub-regions. This helps in providing additional information that could help in improving the classification accuracy. Also, linear subspace approaches such as PCA will not be able to capture the relationship among more than two variables. They cannot depict the variations caused by illuminations, expressions, etc., properly.

In order to capture the relationships among more than two pixels, the data is projected into nonlinear higher dimensional spaces using the kernel method. This enables to capture the nonlinear relationships among the pixels within the modules. Face recognition accuracy in thermal images has been lower compared to visible images under controlled environments. The major problems in thermal face recognition are

(31)

(ii) Opaqueness of glass and, hence, partially occluded faces.

Thermal imaging is sensitive to temperature changes in the surrounding environment. Also it is sensitive to variations in the heat patterns of the face. Factors that could contribute to these variations include facial expressions, physical conditions such as lack of sleep, and psychological conditions such as fear, stress, excitement, etc. Although the problems faced by thermal facial recognition are different from the visual counterpart, they are similar in the sense that many of the variations are confined to local regions. The proposed feature selection process overcomes all the above mentioned challenges due to its robustness to variations caused due to facial expressions, illumination, contrast and partial occlusions.

(32)

CHAPTER 4

(33)

CHAPTER 4 IMPLEMENTATION OF PROJECT

4.1 LIVENESS ASSESSMENT IN AUTHENTICATION SYSTEM:

Liveness assessment methods represent a challenging engineering problem as they have to satisfy certain demanding requirements:

(i) Non-invasive, the technique should in no case be harmful for the individual or require an excessive contact with the user;

(ii) User friendly, people should not be reluctant to use it;

(iii) Fast results have to be produced in a much reduced interval as the user cannot be asked to interact with the sensor for a long period of time;

(iv) Low cost, a wide use cannot be expected if the cost is excessively high;

(v) Performance, in addition to having a good fake detection rate, the protection scheme should not degrade the recognition performance (i.e., false rejection) of the biometric system.

4.2 SOFTWARE-BASED TECHNIQUES:

In this case the fake trait is detected once the sample has been acquired with a standard sensor (i.e., features used to distinguish between real and fake traits are extracted from the biometric sample, and not from the trait itself).

(34)

multi-attack protection method which is used to overcome these limitations by using image quality assessment (IQA). It is not only capable of operating with a very good performance under different biometric systems (multi-biometric) and for diverse spoofing scenarios, but it also provides a very good level of protection against certain non-spoofing attacks (multi-attack).

Moreover, the software based techniques for this type of systems provides more advantages and fast and accurate responses, as it only needs one image (i.e., the same sample acquired for biometric recognition) to detect whether it is real or fake. And this technique is user-friendly (transparent to the user) and doesn‘t allow the fake users; cheap and easy to embed in already functional systems (as no new piece of hardware is required).

An added advantage of the proposed technique is its speed and very low complexity, which makes it very well suited to operate on real scenarios (one of the desired characteristics of this type of methods). As it does not deploy any trait-specific property (e.g., minutiae points, iris position or face detection), the computation load needed for image processing purposes is very reduced, using only

general image quality measures fast to compute, combined with very simple

classifiers. It has been tested on publicly available attack databases of iris, fingerprint and 2D face, where it has reached results fully comparable to those obtained on the same databases and following the same experimental protocols by more complex trait-specific top-ranked approaches from the state-of-the-art.

4.3 IMAGE QUALITY ASSESSMENT FOR LIVENESS DETECTION

:

The use of image quality assessment for Liveness detection is motivated by the assumption that: ―It is expected that a fake image captured in an attack attempt will have different quality than a real sample acquired in the normal operation scenario for which the sensor was designed.‖ Expected quality differences between real and fake samples may include: degree of sharpness, color and luminance levels, local artifacts, amount of information found in both type of images (entropy), structural distortions or natural appearance.

For example, iris images captured from a printed PROJECT are more likely to be blurred or out of focus due to trembling; face images captured from a mobile device will probably be over-or under-exposed; and it is not rare that fingerprint images captured from a gummy finger present local acquisition artifacts such as spots and patches. Furthermore, in an eventual attack in which a synthetically

(35)

produced image is directly injected to the communication channel before the feature extractor, this fake sample will most likely lack some of the properties found in natural images. Following this ―quality-difference hypothesis, in the present research work we explore the potential of general image quality assessment as a protection method against different biometric attacks (with special attention to spoofing). As the implemented features do not evaluate any specific property of a given biometric modality or of a specific attack, they may be computed on any image. This gives the proposed method a new multi-biometric dimension which is not found in previously described protection schemes. In the current state-of-the-art, the rationale behind the use of IQA features for liveness detection is supported by three factors:

Image quality has been successfully used in previous works for image manipulation detection and stag analysis in the forensic field. To a certain extent, many spoofing attacks, especially those which involve taking a picture of a facial image displayed in a2D device (e.g., spoofing attacks with printed iris or face images), may be regarded as a type of image manipulation which can be effectively detected, as shown in the present research work, by the use of different quality features.

In addition to the previous studies in the forensic area, different features measuring trait-specific quality properties have already been used for Liveness detection purposes in fingerprint and iris applications. However, even though these two works give a solid basis to the use of image quality as a protection method in biometric systems, none of them is general. For instance, measuring the ridge and valley frequency may be a good parameter to detect certain fingerprint spoofs, but it cannot be used in iris Liveness detection.

On the other hand, the amount of occlusion of the eye is valid as an iris anti-spoofing mechanism, but will have little use in fake fingerprint detection. This same reasoning can be applied to the vast majority of the Liveness detection methods found in the state-of-the art. Although all of them represent very valuable works which bring insight into the difficult problem of spoofing detection, they fail to generalize to different problems as they are usually designed to work on one specific modality and, in many cases, also to detect one specific type of spoofing attack.

Human observers very often refer to the different appearance‖ of real and fake samples to distinguish between them. As stated above, the different metrics and

(36)

methods designed for IQA intend to estimate in an objective and reliable way the perceived appearance of images by humans. Different quality measures present different sensitivities to image artifacts and distortions. For instance, measures like the mean squared error respond more to additive noise, whereas others such as the spectral phase error are more sensitive to blur; while gradient-related features react to distortions concentrated around edges and textures.

Therefore, using a wide range of IQMs exploiting complementary image quality properties should permit to detect the aforementioned quality differences between real and fake samples expected to be found in many attack attempts (i.e., providing the method with multi- attack protection capabilities). All these observations lead us to believe that there is sound proof for the quality-difference‖ hypothesis and that image quality measures have the potential to achieve success in biometric protection tasks.

4.4 FULL-REFERENCE IQ MEASURES:

Full-reference (FR) IQA methods rely on the availability of a clean undistorted reference image to estimate the quality of the test sample. In the problem of fake detection addressed in this work such a reference image is unknown, as the detection system only has access to the input sample. In order to circumvent this limitation, the same strategy already successfully used for

(37)

image manipulation detection and for stage analysis, is implemented here. The input grey-scale image I (of size N × M) is filtered with a low-pass Gaussian kernel (σ = 0.5 and size 3 × 3) in order to generate a smoothed version Î .Then, the quality between both images (I and Î) is computed according to the corresponding full-reference IQA metric. This approach assumes that the loss of quality produced by Gaussian filtering differs between real and fake biometric samples.

Gaussian filter:

In electronics and signal processing, a Gaussian filter is a filter whose impulse response is a Gaussian function (or an approximation to it). Gaussian filters have the properties of having no overshoot to a step function input while minimizing the rise and fall time. This behavior is closely connected to the fact that the Gaussian filter has the minimum possible group delay. It is considered the ideal time domain filter, just as the since is the ideal frequency domain filter. These properties are important in areas such as oscilloscope and digital telecommunication systems.

Mathematically, a Gaussian filter modifies the input signal by convolution with a Gaussian function; this transformation is also known as the Weierstrass transform. The Gaussian function is non-zero for x \in (-\infty,\infty) and would theoretically require an infinite window length. However, since it decays rapidly, it is often reasonable to truncate the filter window and implement the filter directly for narrow windows, in effect by using a simple rectangular window function. In other cases, the truncation may introduce significant errors. Better results can be achieved by instead using a different window function; see scale space implementation for details.

Filtering involves convolution, the filter function is said to be the kernel of an integral transform. The Gaussian kernel is continuous. Most commonly, the discrete equivalent is the sampled Gaussian kernel that is produced by sampling points from the continuous Gaussian. An alternate method is to use the discrete Gaussian kernel which has superior characteristics for some purposes. Unlike the sampled Gaussian kernel, the discrete Gaussian kernel is the solution to the discrete diffusion equation. Since the Fourier transform of the Gaussian function yields a Gaussian function, the signal (preferably after being divided into overlapping windowed blocks) can be transformed with a Fast Fourier transform, multiplied with a Gaussian function and transformed back. This is the standard procedure of applying an arbitrary finite

(38)

impulse response filter, with the only difference that the Fourier transform of the filter window is explicitly known. Due to the central limit theorem, the Gaussian can be approximated by several runs of a very simple filter such as the moving average. The simple moving average corresponds to convolution with the constant B-spline ( a rectangular pulse ), and, for example, four iterations of a moving average yields a cubic B-spline as filter window which approximates the Gaussian quite well.

4.4.1 Different types FR-IQ Measurements:

Error Sensitivity Measures: Traditional perceptual image quality assessment approaches are based on measuring the errors (i.e., signal differences) between the distorted and the reference images, and attempt to quantify these errors in a way that simulates human visual error sensitivity features. Although their efficiency as signal fidelity measures is somewhat controversial, up to date, these are probably the most widely used methods for IQA as they conveniently make use of many known psychophysical features of the human visual system, they are easy to calculate and usually have very low computational complexity. Several of these metrics have been included in the 25-feature parameterization proposed in the present work. For clarity, these features have been classified here into five different methods those are listed below:

 Pixel Difference measures: These features compute the distortion between two images on the basis of their pixel wise differences. Here we include: Mean Squared Error (MSE), Peak Signal to Noise Ratio (PSNR), Signal to Noise Ratio (SNR), Structural Content (SC), Maximum Difference (MD), Average Difference (AD), Normalized Absolute Error (NAE), R-Averaged Maximum Difference (RAMD) and Laplacian Mean Squared Error (LMSE). The formal definitions for each of these features are given in Table I. In the RAMD entry in Table I, maxr is defined as the r -highest pixel difference between two images. For the present implementation, R = 10. In the LMSE entry in Table I, h(Ii, j ) = Ii+1, j +Ii−1, j + Ii, j+1 + Ii, j−1 − 4Ii, j . 



 Correlation-based measures: The similarity between two digital images can also be quantified in terms of the correlation function. A variant of correlation based measures can be obtained by considering the statistics of the angles between the pixel vectors of the original and distorted images. These features include (also defined in Table I): Normalized Cross-Correlation (NXC), Mean Angle Similarity (MAS) and Mean Angle-Magnitude Similarity

(39)

(MAMS).  

 Edge-based measures: Edges and other two-dimensional features such as corners, are some of the most informative parts of an image, which play a key role in the human visual system and in many computer vision algorithms including quality assessment application . Since the structural distortion of an image is tightly linked with its edge degradation, here we have considered two edge-related quality measures: Total Edge Difference (TED) and Total Corner Difference (TCD). In order to implement both features, which are computed according to the corresponding expressions given in Table I, we use: (i) the Sobel operator to build the binary edge maps IE and ÎE ; ( ii) the Harris corner detector to compute the number of corners Ncr and N̂cr found in I and Î. 



 Spectral distance measures: The Fourier transform is another traditional image processing tool which has been applied to the field of image quality assessment. In this work we will consider as IQ spectral-related features: the Spectral Magnitude Error (SME) and the Spectral Phase Error (SPE), defined in Table I (where F and F̂ are the respective Fourier transforms of I and Î), and arg (F) denotes phase. 

 Gradient-based measures: Gradients convey important visual information which can be of great use for quality assessment. Many of the distortions that can affect an image are reflected by a change in its gradient. Therefore, using such information, structural and contrast changes can be effectively captured. Two simple gradient-based features are included in the biometric protection system proposed in the present article: Gradient Magnitude Error (GME) and Gradient Phase Error (GPE), defined in Table I (where G and Ĝ are the gradient maps of I and Î defined as G = Gx+iGy, where Gx and Gy are the  gradients in the x and y directions).

Structural Similarity Measures:

Although being very convenient and widely used, the aforementioned image quality metrics based on error sensitivity present several problems which are evidenced by their mismatch (in many cases) with subjective human-based quality scoring system. In this scenario, a recent new paradigm for image quality assessment based on structural similarity was proposed following the hypothesis that the human visual system is highly adapted for extracting structural information from the

(40)

viewing field. Therefore, distortions in an image that come from variations in lighting, such as contrast or brightness changes (nonstructural distortions), should be treated differently from structural ones. Among these recent objective perceptual measures, the Structural Similarity Index Measure (SSIM), has the simplest formulation and has gained widespread popularity in a broad range of practical applications. In view of its very attractive properties, the SSIM has been included in the feature parameterization.

Information Theoretic Measures:

The quality assessment problem may also be understood, from an information theory perspective, as an information-fidelity problem (rather than a signal-fidelity problem). The core idea behind these approaches is that an image source communicates to a receiver through a channel that limits the amount of information that could flow through it, thereby introducing distortions. The goal is to relate the visual quality of the test image to the amount of information shared between the test and the reference signals, or more precisely, the mutual information between them. Under this general framework, image quality measures based on information fidelity exploit the (in some cases imprecise) relationship between statistical image information and visual quality. In the present work we consider two of these information theoretic features: the Visual Information Fidelity (VIF) and the Reduced Reference Entropic Difference index (RRED). Both metrics are based on the information theoretic perspective of IQA but each of them takes either a global or a local approximation to the problem, as is explained below.

The VIF metric measures the quality fidelity as the ratio between the total information (measured in terms of entropy) ideally extracted by the brain from the whole distorted image and the total information conveyed within the complete reference image. This metric relies on the assumption that natural images of perfect quality, in the absence of any distortions, pass through the human visual system (HVS) of an observer before entering the brain, which extracts cognitive information from it. For distorted images, it is hypothesized that the reference signal has passed through another ―distortion channel before entering the HVS.

The VIF measure is derived from the ratio of two mutual information quantities: the mutual information between the input and the output of the HVS channel when no distortion channel is present (i.e., reference image information) and the mutual information between the input of the distortion channel and the output of

(41)

the HVS channel for the test image. Therefore, to compute the VIF metric, the entire reference image is required as quality is assessed on a global basis. On the other hand, the RRED metric approaches the problem of QA from the perspective of measuring the amount of local information difference between the reference image and the projection of the distorted image onto the space of natural images, for a given sub-band of the wavelet domain.

In essence, the RRED algorithm computes the average difference between scaled local entropies of wavelet coefficients of reference and projected distorted images in a distributed fashion. This way, contrary to the VIF feature, for the RRED it is not necessary to have access the entire reference image but only to a reduced part of its information (i.e., quality is computed locally). This required information can even be reduced to only one single scalar in case all the scaled entropy terms in the selected wavelet sub-band are considered in one single block.

4.5 NO-REFERENCE IQ MEASURES:

Unlike the objective reference IQA methods, in general the human visual system does not require of a reference sample to determine the quality level of an image. Following this same principle, automatic no-reference image quality assessment (NR-IQA) algorithms try to handle the very complex and challenging problem of assessing the visual quality of images, in the absence of a reference. Presently, NR-IQA methods generally estimate the quality of the test image according to some pre-trained statistical models. Depending on the images used to train this model and on the a priori knowledge required, the methods are coarsely divided into one of three trends:

• Distortion-specific approaches: These techniques rely on previously acquired knowledge about the type of visual quality loss caused by a specific distortion. The final quality measure is computed according to a model trained on clean images and on images affected by this particular distortion.

Two of these measures have been included in the biometric protection method proposed in the present work. The JPEG Quality Index (JQI), which evaluates the quality in images affected by the usual block artifacts found in many compression algorithms running at low bit rates such as the JPEG. The High-Low Frequency Index (HLFI), which is formally, defined in Table.

It was inspired by previous work which considered local gradients as a blind metric to detect blur and noise. Similarly, the HLFI feature is sensitive to the