VEIN IDENTIFICATION

RESEARCH PAPER MADE BY RISHABH SHARMA, ROHIT KUMAR, VIBHOR VARSHNEY

Email:rohitadventure9719@gmail.com,rohitkumarpatel75@gmail.com,omvarsh@gmail.com

Abstract:The palm vein pattern is unique biometric identity of the human beings. The palm vein recognition is a popular biometric technique which is used for authentication purposes in various applications. In this paper a review on new novel palm vein recognition architecture has been described. All of the techniques used to build the palm vein recognition architecture have been discussed in this paper. A palm vein recognition system consists of the following steps: Image acquisition from the database and Pre-Processing, Finding of Region of interest, Extraction of Palm Vein pattern Features and Recognition. The aim of the proposed model is to improve the accuracy and response time of palm vein authentication. The proposed model will use neural networks for the final evaluation of the testing sample and training samples to recognize the person. The proposed model will be robust, flexible and accurate than the existing palm vein authentication schemes.Palm vein authentication is one of the most reliable authentication. Which contain physiological body characteristics that can be used to distinguish between users. In this paper we represent in new approach for the personal identification in the palm vein image.This paper represent the contactless palm vein authentication device that takes blood vessel pattern as a personal identification. It is one of the most critical and challenging task to meet upcoming demand for stringent security. Vein are internal in the body and have wealth of differentiating feature assuming false identity through forgery is externally difficult, thereby enabling and externally high level of security. The palm secure works by capturing a person’s vein pattern image while radiating it with near infrared rays. The palm secure detects the structure of the pattern of veins on palm of the human hand with the at most precision .The sensor emits a near infrared beam towards the palm of the hand and the blood flowing through these back to the heart with reduced oxygen absorb as black pattern. This pattern is a recorded by the sensor and is stored in encrypted from in a database, on a token or an a smart card. It is consist of small palm vein scanner. That is easy and nature to use, fast and highly accurate.

INTRODUCTIONIn the ubiquitous network society, where individuals caneasily access their information anytime and anywhere, people are also faced with the risk that others can easily access the same information anytime and anywhere. Because of this risk, personal identification technology is used which includesPasswords, personal identification numbers and identification cards. However,8cards canbestolenand passwords andnumbers can be guessed or forgotten. To solve these problems, four methods are developed: fingerprints, faces, voice prints and palm veins. Among these, because of its high accuracy, contact less palm vein authentication technology is being incorporated into various financial solution products for use in public places. Palm vein authentication is one of the vascular patternauthentication technologies Vascular pattern authentication includes vein patternauthentication includes vein pattern authentication using the vein patterns of thepalm, back of the hand or fingers as personal identification data, and retina recognition using the vascular patterns at the back of the eye aspersonal identification.

The vascular pattern used in this authentication technology refers to the image of vessels within the body that can be seen as a random mesh at the surface of the body. Since everyone has vessels, vascular pattern authentication can be applied to almost allpeople. If vascular patterns were compared to the features used in other biometric authentication technologies, such as the face, iris, fingerprint, voice, and so on, the only difference would be whether or not the feature is at the surface of the body. Consequently, vascular patterns cannot be stolen by photo graphing, tracing, or recording them. This means that forgery would be extremely difficultunder ordinary conditions.

An individual's palm vein image is converted by algorithms into data points, whichis then compressed, encrypted, and stored by thesoftware and registered along with the otherdetails in his profile as a reference for future comparison. Then, each time a person logs inattempting to gain access by a palm scan to a particular bank account or secured entryway,etc., the newly captured image is likewise processed and compared to the registered one orto the bank of stored files for verification, all in a period of seconds. Numbers and positions ofveins and their crossing points are all compared and, depending on verification, the person iseither granted or denied access.

PALM VEIN TECHNOLOGY Palm vein authentication works by comparingthe pattern of veins in the palm(which appear as blue lines) of a person beingauthenticated with a pattern stored in a database.Vascular patterns are unique to each individual, according toFujitsuresearch even identical twins have different patterns and since the vascular patterns exist inside the body, theycannot be stolen by means of photography,contains a wealth of differentiating features for personal identification. The palm is an ideal partof the body for this technology; it normally does not have hair which can be an obstacle forphotographing the blood vessel pattern, and it isless susceptible to a change in skin color, unlikea finger or the back of a hand.The deoxidized hemoglobin in the vein vessels absorbs light having a wavelength ofabout 7.6 x 10-4 mm within the near-infraredarea. When the infrared ray image is captured, only the blood vessel pattern containing thedeoxidized hemoglobin is visible as a series of dark lines. Based on this feature, the veinauthentication device translates the black linesof the infrared ray image as the blood vesselpattern of the palm, and then matches it with thepreviously registered blood vessel pattern of theindividual.

Principles of vascular pattern authentication:Hemoglobin in the blood is oxygenated in the lungs and carries oxygen to the tissues of the body through the arteries. After it releases its oxygen to the tissues, the deoxidizedhemoglobin returns to the heart through the veins. These two types of hemoglobin have different rates of absorbency1. Deoxidized hemoglobin absorbs light at a wavelength ofabout 760 nm in

the near-infrared region. When the palm is illuminated with near-infrared light, unlike the image seen by the human eye [Figure 1(a)], the deoxidized hemoglobin in the palm veins absorbs this light, thereby reducing the reflection rate and causing the veins to appear as a black pattern [Figure 1(b)]. In vein authentication based on this principle, the region used for authentication is photographed with near-infrared light, and the vein pattern is extracted by image processing [Figure 1(c)] and registered. The vein pattern of the person being authenticated is then verified against the pre-registered pattern.

Working process:

The scanner makes use of a special characteristic of the reduced hemoglobin coursing through the integrated optical system in the palm vein sensor uses this phenomenon to generate animage of the palm vein pattern and the generated image is digitized, encrypted and finally stored as a registered template in the database. The parameters a false rejection rate (FRR) and a false acceptance rate (FAR) is to be calculated for performance measurement. Also,if your profile is registered with your right hand, don't log in with your left, the patterns of an individual's two hands differ. And if you registered your profile as a child, it’ll still be recognized as you grow, as an individual's patterns of veins are established in utero.1. Data from people ranging from 6 to 85 years old including people in various occupations in accordance with the demographics realized by the Statistics Center of the Statistics Bureau.2. Data about foreigners living in Japan in accordance with the world demographics released by the United Nations.

In addition to being contact less and thereby hygienic and user-friendly in that theuser does not need to physically touch a surfaceand is free of such hygiene concerns, palm vein authentication is highly secure in that the veins are internal to the body and carry a wealth of information, thereby being extremely difficult toforge.Also proposed a hierarchical combination scheme for a multimodal identification system. Kittler etc have experimented with several fusion techniques for face and voice biometrics, including sum, product, minimum, median, and maximum rules and they have found that the sum rule outperformed others. Kittler et al. note that the sum rule is not significantly affected by the probability estimation errors and this explains its superiority. Hong and Jain proposed an identification system based on face and fingerprint, where fingerprint matching is applied after pruning the database via face matching. Ben-Jacob et al. considered several fusion strategies, such as support vector machines, tree classifiers and multi-layer perceptron, for face and voice biometrics. The Bayes classifier is found to be the best method. Ross and Jain combined face, fingerprint and hand geometry biometrics with sum, decision tree and linear discriminant-based methods. The authors report that sum rule outperforms others.

Result of experiments:As a result of the Fujitsu research using data from 140,000 palms (70,000 individuals), Fujitsu has confirmed that the FAR is 0.00008% and the FRR is 0.01%, with the following condition: a person must hold the palm over the sensor for three scans during registration, and then only one final scan is permitted to confirm authentication. In addition, the following data has been used to confirm the accuracy of this technology: data from 5-year to 85-year old people of various backgrounds based on statistics from the Ministry of Internal Affairs and Communications of Japan’s population distribution; data from foreigners in Japan based on the world population distribution announced by the U.N.; data of the daily changes of Fujitsu employees tracked over several years; and data of various human activities such as drinking, bathing, going outside, and waking up..

Contact Less Palm Vein Authentication Device:A number of studies showing the advantages of multimodal biometrics fusion have appeared in the literature. BrunelliandFalavigna used hyperbolic tangent for normalization and weighted geometric average for fusion of voice and facebiometrics.They also proposed a hierarchical combination scheme for a multimodal identification system. Kittler et al. have experimented with several fusion techniques for face and voice biometrics, including sum, product,minimum,median, and maximum rules and they have found that the sum rule outperformed others. Kittler et al. note that the sum rule is notsignificantlyaffected by the probability estimation errors and this explains its superiority. Hong and Jain proposed an identification system based on face and fingerprint, where fingerprint matching is applied after pruning the database via face matching. Ben-Yacoub et al. considered several fusion strategies, such as support vector machines, tree classifiers and multi-layer perceptron, for face and voice biometrics. The Bayes classifier is found to be the best method. Ross and Jain combined face, fingerprint and hand geometry biometrics with sum, decision tree and linear discriminant-based methods. The authors report that sum rule outperforms others.

PRE-PROCESSING

Region of Interest Segmentation:

The acquired palm vein images are firstly normalized to minimize the rotational, translation and scale changes. Therefore, to make the identification processeffective and efficient, it is necessary to construct a coordinate system that is invariant/robust to (or nearly) those variations. It is intuitive to associate the coordinate system with the palm itself,and thus the web between index finger and middle finger together with the web between ring finger and little finger were utilized as the reference points/line to build up the coordinate system .

where LROI denotes the side length of ROI, LD denotes the distance between the ROI and the reference line, and LW represents the distance between the two webs, α and β are the factors that control respectively the location and size of the ROI. The present approach is very similar to the method used in [5], however, our computations are simpler since no additional sampling is employed. After this segmentation, ROI is resized to a fixed size (128 × 128 pixels in our experiments) to facilitate the identification processes.

Image Enhancement :Since the palm vein images employedin our work were acquired under near infrared illumination (NIR), the image appears dark. Therefore to achieve better visualization of vascular and surface details, contrast enhancement is required. We firstly estimate the background intensity profiles by dividing the image into slightly overlapped 32 × 32 blocks (3 pixel overlapping between two blocks to address the ‘blocky effect’), and the average values of each block are calculated. Subsequently, the estimated background intensity profile is resized to the same size as the original image using bicubic interpolation and then is subtracted from the original ROI image. Finally, the histogram equalization was employed to obtain the enhanced image. The comparison between figures suggests that the enhancement scheme has been quite successful in enhancing the local surface details.

Block diagram for the personal identification using multiple representations of palm vein images.

Advantages of using the palm:

In addition to the palm, vein authentication can be done using the vascular pattern on the back of the hand or a finger. However, the palm vein pattern is the most complex and covers the widest area. Because the palm has no hair, it is easier to photograph its vascular pattern. The palm also has no significant variations in skin color compared with fingers or the back of the hand, where the color can darken in certain areas.

VEIN PATTERN RECOGNITIONNearly any part of vein in human body (such as retinal vein, facial vein, veins in hand) could be used for personal identification, but veins in hand are always preferred [6]. It is usually an uncovered part. Veins in hand are closer to the surface than other organizes, so the traits can be easier detected by low-resolution cameras. In this paper, vein in hand is involved, finger vein, palm vein, wrist vein and dorsal hand vein, and each of them offers stable and unique biometric features.Finger Vein Authentication System captures images of the vein patterns inside your finger. These, like other biometric patterns are unique – but significantly, because they are inside your body, they are virtually impossible to replicate. The method works by passing near-infrared light through the finger. This is partially absorbed by the hemoglobin in the veins, allowing an image to be recorded on a CCD camera underneath. Then, it takes around 0.5 of a second to match your vein patterns to those previously digitized, compressed and captured on a smart card. The process is remarkably accurate – there’s only a 0.0001% chance of someone passing off their vein patterns as yours, as accurate as established iris systems. And, of course, for additional security, you can always record the vein patterns in more than one finger.

Finger vein authentication process

IMAGING PRINCIPILEAs veins are internal, their structure cannot be discerned in visible light. Based on the kinds of light of acquisition, a vein image can be classified as X-ray scanning, ultrasonic scanning and infrared scanning. X-ray and ultrasonic are used to capture vein images in medical treatment, but they are not used in identification due to the health case. Until now, researchers used infrared imaging for personal identification. Infrared (IR) is electromagnetic radiation whose wavelength is longer than that of visible light, and Infrared light has a range of wavelengths lies between about 750nm and 1mm, just like visible light has wavelengths that range from red light to violet. Infrared is commonly divided into 3 spectral regions: near, mid and far-infrared light, but the boundaries between them are not agreed upon. There are two choices that focuses on imaging of vein patterns in hand by infrared light, the far-infrared (FIR) imaging and the near-infrared (NIR) imaging, which are suitable to capture human bodies images in a non-harmful way. Some papers had discussed the principle of the FIR and NIR imaging methods. In the FIR method, superficial human veins have higher temperature than the surrounding tissues. ForNIR light method, the principle could be explained by photobiology. In biology, there is a “medical spectral window”, which extends approximately from about 740 to 1100 nm. The light in this window could penetrate deeply into tissues. Because blood and surrounding tissues have different effect on the NIR light, we could use a CCD camera with an attached IR filter to capture images in which vein appears darker.

FIR WayThe human body temperature is about 36.85°C, and the temperature of surface of human veins is higher than that of the surrounding parts. Therefore when the FIR light irradiates hand, the hand vein structure is thermally mapped by an infrared camera at room temperature. The captured image shows a gradient of temperature between surrounding tissues and the back-of-hand veins.

In literature , it is proved that the captured FIR image of the back of hand has good quality, which means containing more useful information, but FIR vein image at palm and wrist have poor quality. Whilst this method deeply affects by the humidity and temperature of surrounding, as well as the users’ perspiration does.

NIR WayNear infrared wavelength is between about 700 nm to 1400 nm, and we can use the same observing methods as that used for visible light, except for observation by eye. The NIR light is not thermal. NIR scanning device cannot penetrate very deep under the skin therefore the device will recognize the superficial veins and rarely the deep veins. In the NIR way, the light of specific wavelength is almost completely absorbed by the deoxidized hemoglobin in vein while almost penetrated the oxidized hemoglobin in the arteries. Oxygenated and deoxygenated hemoglobin absorb light equally at 800 nm, whereas at 760 nm absorption is primarily from deoxygenated hemoglobin . Then the veins appear as dark areas in an image taken by a CCD camera. Near-infrared (NIR) spectroscopy is a noninvasive technique that uses the differential absorption properties of hemoglobin to evaluate skeletal muscle oxygenation.NIR method is not a temperature based technique since normal body temperature or surrounding temperature cannot interfere with this method. The FIR method is often used in hand-dorsa vein imaging, and NIR method can be used in all veins imaging in hand. In order to benefit the processing, the captured images are always the grayscale image.

Vein pattern extractionBecause the temperature, illumination, locus and angle vary each collection, the captured digital picture varies each time. In order to provide ‘better’ input for automated image processing and realize a robust system against some fluctuation, some form of normalization should to be done a forehand. Conventional preprocessing algorithms can do this work. Then the vein patterns are extracted after noise reduction and normalization. Several algorithms have been carried out to separate the vein patterns from the image background. The captured images contain shading, noise and vein patterns, moreover, the vein patterns are not salient. The more the information of veins is extracted and preserved, the better the accuracy is. So the appropriate processing extracting the vein patterns is important for the authentication system. Recently vein of hand extraction algorithm has been widely studied. Wherever the veins are, in finger, wrist, palm or the back of hand, the various forms of vein patterns extracting algorithms usually fall into four broad categories: tracking-based, transform-based, matched filter method and thresholding method. Here we will describe some work on each of these areas.

Tracking-basedThe tracing algorithm is based on repeated line tracking the vein from initial seed-point in the captured NIR image, moving pixel by pixel along the dark line in the cross sectional profiles [11]. In figure6, there is a certain position ‘s’, and the left is its cross sectional intensity profile of finger vein image. Tracking direction is determined by the position of deepest point in the cross sectional. This method can extract vein patterns from low quality NIR images, but it is sharply affected by the temporal change of widths of veins.

Transform-based methods:The captured image always has low contrast and contains noise, so contrast enhancement and noise reduction are crucial in ensuring the quality of the subsequent steps. Transform based methods can convert image to a certain domain in which it is more suitable for extracting the patterns. Wavelet, which supports multi-resolution analysis, is one of the appropriate methods for vein structure and feature extracting. The wavelet multi-resolution approach employs a wavelet basis to analyze at different resolutions and increase resolution from coarse to fine, so the content of image in each scale can be understood. Vein patterns are well structured objects consisting of line-like veins and areas in between. The wider veins can be analyzed in the lower resolution, and the thinner veins can be analyzed in the higher resolution. In paper [12], dyadic wavelet transform is adopted to extract finger vein patterns from background. Image is transformed from spatial domain to wavelet domain, and the grayscale image is changed into wavelet coefficients, which contain vein patterns wavelet coefficients and noise wavelet coefficients. The vein pattern variance of coefficients is larger than that of noise, and with the increasing of wavelet scale, the noise variance decreases.

Matched filter method:By observing the cross sectional profiles of vein patterns, some researchers proposed an intensity profile model to detect vein patterns. Several models have been presented to describe the cross sectional profile of vessel [13-15]. The gray-level profile of the cross section is approximated a Gaussian shaped curve, which is prevalent used, whilst the matched filter is utilized to detect vein patterns. Since vein patterns may appear in any orientation, a set of cross sectional profiles in equiangular rotations is employed as a filter bank.

Thresholding method:Intensity thresholding is usually utilized to obtain a better representation of shapes of the vein patterns. In the IR image the different location has different intensity values of the veins. Hence applying a single global thresholding is inappropriate. Via adaptively adjusting local thresholding, we can choose different threshold values for every pixel in the image based on the analysis of its surrounding neighbors [9], then, separate the vein patterns from the background, after that the desired vein image is extracted.

Pattern matching:The extracted vein patterns of the input image can directly be compared with the templates. A certain distance is defined to calculate the similarity between the template and the input patterns. But when the template is not small, the comparing time lasts long. After pattern extracting process, most systems are interesting in eliciting skeletonisation of the vein patterns. Then Vessels can be represented by the number of intersections, the total segment length, the longest segment, and the angles found in the image, the distribution of the vein, and other statistical features. Hausdorff distance, SVM, and nearest neighbor are adopted as matching algorithm by researchers.

Database:Recently, significant work is continuously being done in vein recognition algorithms both in academy and industry. However, the conclusion of each work is usually achieved on their own databases but not the sharable databases. Large sharable vein databases are required to evaluate and compare various algorithms. Vein pattern data collection is an expensive and time-consuming work. There are some inconveniences in large databases collection. Firstly, it is expensive both in terms of money and time; secondly, it is tedious for both the technicians and for the volunteers; thirdly, due to privacy information, it is difficult to share data with others. Though the real images cannot be replaced, the synthetic vein images have proven to be a valid substitute for real vein for design, benchmarking and evaluation of vein recognition systems. A synthetic like-vein image method is requested. Based on the cross sectional profiles of vein patterns, the vein pattern can be synthesized in semiautomatic way as figure10. Firstly, lines which look like vein patterns were drawn by hand. Secondly, according to the different cross sectional profile models, the like-vein patterns can generation by programs.

Application of vein recognition system and future workVein recognition technology has some fundamental advantages over fingerprint systems. Vein patterns in hand are biometric characteristics that are not left behind unintentionally in everyday activities. Vein patterns of inanimate bodily parts become useless after a few minutes. Hence, nowadays, vein recognition system is regarded a mainstream technology. IBG expects it to play a larger role and comprise more than 10% of the biometric market [18]. Nearly all major vein authentications are manufactured in Japan and Korea, and theapplication of these manufactures is used in Asia. In Japan and some other countries, suchproducts spread particularly in the financial sector.

Fig: (a) Finger Vein device; b) Finger Vein ATM; c) Palm Secure by Fujitsu

The recent launch of vein recognition technology is successful. Nevertheless, some research issues need to be addressed in future. For one thing, work continued across the vein imaging device to make it cheaper, more accurate and robust. For another thing, the quality of vein IR image is affected by the relationship of intensity between the IR light and the ambient light, as well as the ambient temperature. Moreover, the sharable large databases should be founded for a thorough evaluation on the efficacy of different vein recognition algorithms. Lastly, vein trait is able to conjunct with other biometrics in a multi-modal system.

MARKETINGA reliable biometric system, which is essentially a pattern-recognition that recognizes a person based on physiological or behavioral characteristic , is an indispensable element in several areas, including ecommerce(e.g. online banking), various forms of access control security(e.g. PC login), and so on. Nowadays, security has been important for privacy protection and country in many situations, and the biometric technology is becoming the base approach to solve the increasing crime.As the significant advances in computer processing, the automated authentication techniques using various biometric features have become available over the last few decades. Biometric characteristics include fingerprint, face, hand/finger geometry, iris, retina, signature, gait, voice, hand vein, odor or the DNA information, while fingerprint, face, iris and signature are considered as traditional ones. IBG Biometric Market by Technology Due to each biometric technology has its merits and short coming, it is difficult to make a comparison directly. Jain et al. have identified seven factors, which are (1) universality, (2) uniqueness, (3) permanence, (4) measurability, (5) performance, (6) acceptability, (7) circumvention, to determine the suitability of a trait to be used in a biometric application.

Vein pattern is the network of blood vessels beneath person’s skin. The idea using vein patterns as a form of biometric technology was first proposed in 1992, while researches only paid attentions to vein authentication in last ten years. Vein patterns are sufficiently different across individuals, and they are stable unaffected by ageing and no significant changed in

adults by observing. It is believed that the patterns of blood vein are unique to every individual, even among twins. Contrasting with other biometric traits, such as face or fingerprint, vein patterns provide a really specific that they are hidden inside of human body distinguishing them from other forms, which are captured externally. Veins are internal, thus this characteristic makes the systems highly secure, and they are not been affected by the situation of the outer skin (e.g. dirty hand). At the same time, vein patterns can be acquired by infrared devices by two ways, noncontact type and contact type. In the case of non-contact method, there is no need to touch the device, and therefore it is friendly to individuals in the target population who utilize the systems. In the contact type, the collection type is the same as fingerprint which has already been accepted by most people. From the customer’s point of view, the authentication system is not only high accuracy level for security but also easy to enroll. Vein patterns serve as a high secure form of personal authentication as iris recognition (Iris is known for high accurate rates of authentication, but it is regarded unfriendly by users due to the direct application of light into their eyes), and serve as a convenient form as fingerprint recognition. On account of the several advantages, vein authentication is not only interested in lab researchers but also in industries, and the products perform well in tests of the International Biometric Group (IBG). Recently, vein recognition appears to be making real headway in the market, and considered as one of the more ’novel’ biometric, which is called ‘the Fourth ‘Biometric’.

REFERENCES.Prabhakar, S.Pankanti, and A. K. Jain, “Biometric Recognition: Security and Privacy Concerns”, IEEE Security and Privacy, 2003.1(2), pp. 33-42.J. L. Wayman, A. K. Jain, D. Maltoni, and D. Maio, “Biometric Systems: Technology, Design and Performance Evaluation”, 2005, Springer.International Biometric Group, “Biometrics Market and Industry Report 2007-2012”, 2007. A. K. Jain, R. Bolle, and S. Pankanti, “Biometrics: Personal Identification in Networked Society”, 1999, Kluwer Academic Publishers.Michael Thieme, “New: Vein performs well in tests”, Biometric Technology Today, 2006.14(10), pp. 4.S. Crisan, l. G. Tarnovan, and T. E. Crisan, “A Low Cost Vein Detection System Using Near Infrared Radiation”, IEEE Sensors Applications Symposium 2007, San Diego, California USA, 2007.A. K. Jain, A. Ross, and S. Prabhakar, "An Introduction to Biometric Recognition", IEEE Trans. on Circuits and Systems for Video Technology, 2004.14(1), pp. 4-19.J. Hashimoto, “Finger Vein Authentication Technology and its Future”, VLSI Circuits, 2006.Digest of Technical Papers. 2006 Symposium on, 2006, pp. 5-8.L. Wang, G. Leedham and S. Y. Cho, “Infrared imaging of hand vein patterns for biometric purposes”, The Institution of Engineering and Technology 2007 IET Comput Vis., 2007, pp. 113–122.D. M. Mancini, L. Bolinger, H. Li, K. Kendrick, B. Chance and J. R. Wilson, “Validation of near-infrared spectroscopy in humans”, Journal of Applied Physiology, 1994. 77(6), pp. 2740-2747.Miura, N., A. Nagasaka, and T. Miyatake, “Feature extraction of finger-vein patterns based on repeated line tracking and its application to personal identification”, Machine Vision and Applications, 2004.15(4), pp. 194-203.

Li, Xueyan, Guo, Shuxu, Gao, Fengli, and Li, Ye, “Vein Pattern Recognitions by Moment Invariants”, The 1st International Conference on Bioinformatics and BiomedicalEngineering, 2007, pp. 612-615.A. Hoover, V. Kouznetsova, and M. Goldbaum, “Locating blood vessels in retinal images by piece-wise threshold probing of a matched filter response”, IEEE Trans. on Medical Imaging, 2000.19(3), pp. 203–210.L. Gang, O. Chutatape, and S. M. Krishnan, “Detection and Measurement of Retinal Vessels in Fundus Images Using Amplitude Modified Second-Order Gaussian Filter”, IEEE Trans on Biomedical Engineering, 2003. 49(2), 168-172.