Algorithm Development for Fingerprint Image Enhancement Using Wavelet Processing

Algorithm Development for Fingerprint Image Enhancement Using Wavelet Processing

Chapter One

Preamble of the Study

Today’s world has been greatly modified by technology in the efforts of man to deal with his environment in order to tame it to suit his needs; with particular reference to electronics. Virtually, all aspects of our modern civilization depend on electronic [including hardware and software] systems for smooth and effective operations of daily activities. Ranging from military, communications, government, education, banking, automobiles and manufacturing to the space industry; all rely on electronic systems for various applications like process automation, control and monitoring systems, security, automated parking control, 24 hours surveillance systems, traffic collision avoidance system (TCAS) of aircrafts and on board data handling subsystems of satellites (OBDH). With the dynamic growth of the world’s population, global economic meltdown, food scarcity, and unemployment; all these factors have led to significant increase in violent crimes globally making security a great challenge to all and sundry.

Governments, business owners, the organized private sector and educational institutions all require a robust, rugged and dependable means of recognizing and verifying peoples’ identities at various points of access to specific areas physically and electronically.

In order to ensure that intruders, impersonators are promptly fished out prior to breaching the security of these valued resources, information, assets and vaults; a robust and effective means of identifying and verifying peoples’ identities is paramount. The search for such systems culminated recently in biometrics.

CHAPTER TWO

LITERATURE REVIEW

  PRELIMINARY

An image is defined as a signal of two dimensions that has its intensity specified by a function of two variables in space at any given point [12].

Fingerprint is an impression that is created when the finger(s) is pressed against a flat [or any] surface. They are inherently unique to each human; fingerprints are formed on the fingers of the foetus at about six months and remain unchanged until decomposition after death.

Image enhancement techniques abound under which fingerprint image enhancement has its context. The objective is to improve recognition rate while reducing the rejection rate of enrolled users. Literature abound on image enhancement techniques, in this chapter we will review some state of the art technologies that have been employed to enhance fingerprint images. The main focus will be on wavelet processing techniques applied in fingerprint images enhancement, as this is the primary goal of this research work.

As stated earlier, biometric systems automatically identify and/or verify humans based on their unique inherent characteristics of which fingerprint is the most popular. Methods and techniques that extract these unique characteristics from fingerprint images such minutiae extraction form the foundation of the biometric system.

This review will provide insight on the methodology, overview of some fingerprint image enhancement techniques, summarize works that have been done in this area, evaluate such works as well as provide the context of this thesis. It also aims to identify gaps where there are rooms for improvements and develop understanding of the theories and concepts associated with fingerprint image enhancement algorithms.

Image enhancement techniques select a particular characteristic of an image to adjust the appearance of any given image. The overall objective is to improve the image quality by manipulating its features such as: enhancement of contrast, normalization of illumination, zooming, detection of edges, red eye reduction, and manipulation of shadows. An enhanced image provides detailed information from the original image signals by removing or reducing noise and its effects, this is usually done by applying digital signal processing techniques such as filtering and decomposition. Other methods employed in image quality enhancement include: spectral methods, wave-front or shock methods, use of non-Albelian group operations [i.e. logical image operations and connected operations].

Wavelet processing is the application of mathematical functions that cut up data [for instance an image signal] into different frequency components, and investigates individual parts of the frequency with a corresponding resolution to its own standard or scale for image processing. In order to define a given function as a wavelet system, it must consist of the following three characteristics as stipulated in [13]:

1. General functions were derived with wavelets as basic components or blocks i.e. a particular function or signal is denoted in the wavelet domain as an average of infinite series of
2. There is a space-frequency localization of wavelets implying that majority or a significant portion of the energy in a wavelet is restricted in a fixed range or interval and also frequency components of the transform is within a particular frequency

• The wavelets are capable of handling fast as well as efficient transform algorithms [13].

Since there are numerous image enhancement techniques, this review will be restricted to and organized as follows:

1. Fingerprint image enhancement
2. Wavelet

• Image enhancement using wavelet

1. Fingerprint image enhancement using wavelet

FINGERPRINT IMAGE ENHANCEMENT TECHNIQUES

In [14], a fingerprint was defined as ‘a characteristic decorative pattern in a person’s fingers and skins, which consists of rise lines of skins. The start point, end point, bifurcation, and bonding point, of these rise lines is referred to as fingerprint’s specific character’.

[13] Described fingerprint image enhancement as the next phase that succeeds image segmentation [this process distinguishes image foreground from its background] that is employed to improve image quality. This process clearly separates ridges and valleys of the fingerprint image by enhancing the visual qualities of the image through texture filtering, as well as improvement of the local features of the fingerprint image such as rate and direction.

From [15], we learn that most algorithms deployed towards fingerprint image enhancement that reduce noise while effectively sharpening the contrast of valleys and ridges; rely on gray scale images. These algorithms can be broadly classified into ‘spatial domain filtering enhancement and transformed domain enhancement techniques’ [15]. Examples include Fourier and wavelet domains respectively. According to Yuan et al [15], it is computationally cumbersome to implement spatial domain enhancement techniques as they apply spatial convolution of images during filtering.

[16] And [17] both show that Gabor filters are widely adopted in the areas of fingerprint classification and verification to improve the applications’ performance. Gabor filter being bandpass filters possess frequency as well as orientation selective properties, making them suitable in both frequency and spatial domains for optimal joint resolution [15]. A technique was proposed in [15] in which super resolution is incorporated as a pre- processing step in the enhancement algorithm particularly for low quality images usually obtained from a camera. ‘Super resolution is a technique of enhancing image resolution by combining information from multiple images’ [15]. Recent methods applied in super resolution enhancement of fingerprint images are primarily based on spatial domain. [18] Put forward a proposal that recommended the use of a method that generates high resolution images through a combination of both space and time domain information.

• Suggested a technique called Projection Onto Convex Set (POCS) for super resolution which is currently one of the most applied super resolution techniques. In [20], an extension on the discussion of POCS with modifications aimed at its improvements was made. Another conventional super resolution technique is the stochastic based super resolution, such as that presented in [21] advocated the use of Bayesian estimation in super resolution techniques. Maxima-a-Posteriori (MAP) a common stochastic process was employed in the technique put forward by[22].

₦3,000.00 – DOWNLOAD CHAPTERS 1-5 PROCEED TO DOWNLOAD Added to cart

 

CHAPTER THREE 

RESEARCH METHODOLOGY

In this chapter we will discuss the research methodology, develop fingerprint image enhancement using wavelet processing and present background information on The research methodology consists of a series of steps thatinclude:

1. 1. Input Fingerprint Image: The FVC 2002 set B database was chosen as the source of the input fingerprint image as they are internationally available and recognized for fingerprint algorithm
   2. Normalization: The input fingerprint image is normalized here in order to obtain an evenly illuminated fingerprint image. This is achieved by adjusting the mean and variance values of the matrix generated from the fingerprint

• Decomposition: Daubechies’ wavelet is employed to compute the wavelet transform of the normalized fingerprint image. This mathematical operation decomposes the normalized image into four sub-images namely: approximation detail, vertical detail, horizontal detail and diagonal

1. Reconstruction: The inverse wavelet transform of the image is computed using Daubechies’ wavelet too. This step generates an enhanced image for
2. Wavelet Filtering: Here, the reconstructed fingerprint image is filtered by wavelet means to smooth the enhanced
3. Display enhanced image: This last phase displays the enhanced fingerprint image which can be used for identification or verification

The block diagram of figure 11a shows an overview of the methodology.

CHAPTER FOUR

EXPERIMENTS

DISCUSSION OF EXPERIMENTS ANDPROCEDURE

STEP 1: Downloaded DB1_B of FVC- 2002 as the source for the input fingerprint images for testing the algorithm.

STEP 2: This database comprises of 80 images out of which 8 images where taken at random for experimentation.

STEP 3: The fingerprint images are generally of size 388 x 374 pixels with about 96dpi resolution.

STEP 4: The experimental results produced from the algorithm are shown in tables

1 and 2. The application of wavelet transform to fingerprint image enhancement, offers a robust technique that allows automatic fingerprint verification and identification systems improve their recognition rate. Daubechies’ wavelets where employed, as the sub details resulting from the decomposition contain enough information of the fingerprint image that lead to satisfactory reconstruction of the enhanced fingerprint image. Although it is theoretically possible to decompose the fingerprint image at any level, it was restricted to level 2 in this work as it  was observed that the ridge patterns and valley structures vanished at levels greater than 2 due to excessive down-sampling of the fingerprint image.

Wavelets also offer a very suitable image denoising feature that removes image noise without altering the image quality (i.e. does not compromise the integrity of the fingerprint ridge and valley structure); making it suitable for compression and consequently for transmission applications.

Another great feature of wavelets is their capability for multiresolution processing of images. The scale of the wavelet is proportional to the resolution; hence permitting us to choose any adequate scale that suits our resolution requirements. Wavelets are robust indeed as they provide information of the image signal in both spatial and frequency domains.

CHAPTER FIVE

 RECOMMENDATIONS

From the results of the experiments with the algorithm, it is recommended that fingerprint images be enhanced prior to minutiae extraction in order to increase the recognition rate and consequently reduce false rejection rate (FRR).

For future works in this area, it is recommended that Gabor filtering be included as a step in the enhancement process. This should be inserted after the wavelet decomposition phase before the computation of the inverse wavelet transform. The reason for this is that Gabor filters possess both frequency and orientation selective properties that will preserve the ridge and valley patterns of the sub-detail images aiding a better reconstruction of the final enhanced image. Wavelet denoising approach was found to diminish the ridge and valley structures after decomposition. Hence, the inverse wavelet transform of the enhanced image was computed prior to wavelet denoising.

As a prerequisite for further research in biometrics, the scholar should possess a working knowledge of the MATLAB programming language, as this is the primary tool for implementing most fingerprint image enhancement algorithms.

Set A of FVC 2002 or FVC 2004 or FVC 2006 database should be employed in future works for input fingerprint images to be enhanced as they contain updated fingerprint images from recent advancements in fingerprint image capturing technologies.

For biometric systems to mitigate the various limitations mentioned previously, the following steps can be taken:

1. For most commercial deployment of biometric systems, passwords or tokens can be integrated with biometrics to address theft of fingerprints of authentic users.
2. In high risk and sensitive areas such as military and power plant installations, each component of the biometric system should implement a standalone security measure. When coupled, the components should offer additional

• Design and implementation of vitality detection mechanism (to check for signs of life such as pulse and/papillary hippus) [9] which can be integrated into the biometric system’s software and

1. Perhaps the most formidable approach to biometric security concerns is to design a biometric system that applies two or more biometric features for recognition such as fingerprints, gait, face and voiceprints. These biometric systems are referred to as multimodal biometric
2. A unimodal biometric system based on fingerprint can improve its security by requiring the enrollment of several fingerprints and request that they are provided in a specific order during

CONCLUSION

In this thesis, an effective and efficient fingerprint image enhancement algorithm that is based on wavelet processing has been developed. The algorithm reads an input fingerprint image of uneven illumination, normalizes the illumination, decomposes it into approximation, horizontal, vertical and diagonal sub details using Daubechies’ wavelets. Its inverse wavelet transform was computed and finally denoised the enhanced fingerprint image using wavelet techniques.

The algorithm provides high performance results as can be seen from section 4.4.

References:

• Evans, “Criminal Investigations: Crime Scene Investigation”, J. L. French (Ed), InfoBase Publishing, New York, 2009, pp13-29.
• Ribaric and I. Fratic, “A Biometric Identification System Based on EigenPalm and EigenFinger Features”, IEEE transactions on Pattern Analysis and Machine Intelligence, Vol. 27, № 11, November 2005, p.1698.
• Ma, Y. Hong, and T. Tan, “Iris Recognition Using Circular Symmetric Filters”, IEEE Proceedings, 16^thInt. Conf. on Pattern Recognition, 2002, pp. 414-417.
• Cappelli, A. Lumini, D. Maio and D. Maltoni, “Fingerprint Image Reconstruction from Standard Templates”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 29, № 9, September 2007, p.1489.
• Bansal, P. Arora, M. Gaur, P. Sehgal, and P. Bedi, “Fingerprint Image Enhancement Using Type-2 Fuzzy Sets”, IEEE Sixth Int. Conf. on Fuzzy Systems and Knowledge Discovery, 2009, pp412-417.
• Wang, N. Bhattacharjee, and B. Srinivasan, “Fingerprint Reference Point Detection Based on Local Ridge Orientation Patterns of Fingerprints”, IEEE World Congress on Computational Intelligence,2012.

₦3,000.00 – DOWNLOAD CHAPTERS 1-5 PROCEED TO DOWNLOAD Added to cart