Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Design an Advance computer-aided tool for Image Authentication and Classification

Design an Advance computer-aided tool for Image Authentication and Classification In/SN Date: 30-05-2012 Time: 1: 30 Design an Advance Computer-Aided Tool for Image Authentication and Classification 1.2 1 Rozita Teymourzadeh, Amirize Alpha Laadi, 2 2 2 Yazan Samir Algnabi, MD Shabul Islam and Masuri Othman Department of Electrical and Electronic, Faculty of Engineering, Technology and Built Environment UCSI University, Jalan Choo Lip Kung, Taman Taynton View, 56000 Cheras, Kuala Lumpur, Malaysia Department of VLSI Design, Institute of Micro Engineering and Nanoelectronics IMEN, Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor, Malaysia Abstract: Problem statement: Over the years, advancements in the fields of digital image processing and artificial intelligence have been applied in solving many real-life problems. This could be seen in facial image recognition for security systems, identity registrations. Hence a bottleneck of identity registration is image processing. Approach: These are carried out in form of image preprocessing, image region extraction by cropping, feature extraction using Principal Component Analysis (PCA) and image compression using Discrete Cosine Transform (DCT). Other processing include filtering and histogram equalization using contrast stretching is performed while enhancing the image as part of the analytical tool. Hence, this research work presents a universal integration image forgery detection analysis tool with image facial recognition using Black Propagation Neural Network (BPNN) processor. The proposed designed tool is a multi-function smart tool with the novel architecture of programmable error goal and light intensity. Furthermore, its advance dual database increases the efficiency for high performance application. Results: With the fact that, the facial image recognition will always, give a matching output or closest possible output image for every input image irrespective of the authenticity, the universal smart GUI tool is proposed and designed to perform image forgery detection with the high accuracy of ±2% error rate. Conclusion: Meanwhile, a novel structure that provides efficient automatic image forgery detection for all input test images for the BPNN recognition is presented. Hence, an input image will be authenticated before being fed into the recognition tool. Key words: Principal Component Analysis (PCA), Discrete Cosine Transform (DCT), Black Propagation Neural Network (BPNN), Local Binary Pattern (LBP) INTRODUCTION 40 numbers with admixtures of Eigen pictures can be characterized a face. Facial image recognition is a bottleneck of image Wu and Haung (1990) took advantage of 24 processing that shows a lot of interrest in past few years measurements in fiducially signs of profile double (Kumar et al., 2007). Move towardes to 2-D face facial image. The proposed system has been reduced by recognition leads to have template-based approach and recognizing performance when more image in the geometric-based approach (Bolme et al., 2003). further database because there are not enough to differentiate complex features are composed of several of the feature features that identify a user. Then, Turk and Penland (1991) introduced the examples. This type of representation leads to an Eigen faces method that achieved a remarkable abstraction from the image pixels. acknowledgement speed, but the speed drastically Sirovich and Kirby (1987) proposed Eigen picture reduced, when against resizing. Wiskott et al. (1997) as economically representation of image in a best used magnitude information and phase information that coordinate system. It was observed that an acceptable is received from Gabor wavelet transformation of the picture of a face could be reconstructed from the face image. For identifying a person, this information is specification of gray levels at 214 pixel locations. compared with all of the information in database However, they showed that in actual construction, only Corresponding Author: Rozita Teymourzadeh, Department of Electrical and Electronic, Faculty of Engineering, Technology and Built Environment, UCSI University, 56000, Cheras Kuala Lumpur Malaysia 1 End time searching for the one who has the same transformation as X΄  T  X, Y Y΄  T  X, Y (1) X Y test image. As earlier stated, the geometric operations that are Gallagher (2005) analyzed a set of images and common with most forgeries are re-scaling, rotation and proposed that image forgery most often includes some skewing. Hence, the importance of detection will be form of geometric transformation. These processes traced wit utilization of affine transformations. were based on a re-sampling and interpolation step. The general equations for affine transformations Later on, Ahonen and Hadid (2006) introduced the are given by: Local Binary Pattern (LBP). This operator was used to X΄  a  a x  a y 0 1 2 measure the texture information of local area of gray (2) Y΄  b  b x  b y 0 1 2 image. The second stage is the interpolation. Here pixel intensity values of the transformed image are assigned The various research work had been completed using low pass interpolation filter. According to the (Ming et al., 2008; Mehryar et al., 2010; Jaeyoung and sampling theory, if the Nyquist criterion is satisfied, the Jun, 2011; Han et al., 2011; Lu et al., 2003) to reduce spectrum F(w) does not overlap in the Fourier domain. the efficiency with variations in illumination, image The original signal f(x) can be reconstructed perfectly size, size of database, percentage error. In addition, from its samples f using the optimal sine interpolation inability to detect altered images for more realistic (Hou and Andrews, 1987). recognition still is challenging task. Combining the derivative theorem with the Hence, this study proposes an efficient passive convolution theorem leads to the conclusion that by evaluation tool for authenticity of input images before convolution of f with derivative kernel D (w), it is k n recognition process stage, furthermore, the adjustment possible to reconstruct the nth derivative of the image of illumination of images using contrast stretching and image histogram equalization techniques will be f(x). The result of interpolated operation of f (x) is performed. In this proposed research work, the quality denoted by Equ. 3; percentage is adjustable in order to achieve high performance (as low as 2% error rate) with a dual 𝑓 (𝑥 ) = f w − k database approach. The test image is reduced to a lower 1-D dimensional vector, which represents D {f }(𝑥 ) = D f w − k = distinguishing features of the test image. The 1-D image vectors are fed into the neural f D {w} − k (3) network, the Euclidean distance of the 1-D image vector is compared to each test database and the closest match is found and outputted. By assuming the constant variance random process, n w then the variance of D {f } which is var {D {f }(x)} n w System design fundamental: The focus of the image as a function of x is given by: authentication stage (Fig. 1) is the presence of the { { } } image re-sampling trace and/or interpolation signal. A 𝑟𝑣𝑎 𝐷 𝑓 (𝑥 ) = 𝑅 {𝑓 }(𝑥 , 𝑥 ) = derivative operator on the variance of the image will 𝜎 D {𝑤 } − k (4) detect these traces. Later, the signal is processed through radon transformation that results in a periodic signal embedded or completely imposed on the image Similarly, the covariance is represented as Equ. 5: spectrum graph (Gallagher, 2005). Hence, interpolated signal and re-sampled signal will be ready to function. 𝑅 {𝑓 }(𝑥 , 𝑥 + 𝜉 ) = 𝜎 D {𝑤 } − k Interpolated signal: There are two main stages in xD {𝑤 } − k (5) geometric transform (Popescu and Farid, 2005; Prasad & Ramakrishnan, 2006). In the first stage, a spatial transformation of the physical re-arrangement of pixels in the image is performed and can be represented by a transformation function T as shown in Equ.1. 2 End time Fig. 1: Proposed process structure Now, by assuming that 𝜗 is an integer, it can be magnitude of Dn{b(x, y)} along specified direction determination by an angle θ. noticed that: The projection is a line integral in a certain direction. This line integral is expressed as: { { }( )} { { }( )} 𝑟𝑣𝑎 𝐷 𝑓 𝑥 = 𝑟𝑣𝑎 𝐷 𝑓 𝑥 + 𝜗𝛥 (6) { } { } 𝜌𝐷 𝑏 (𝑥 , 𝑦 ) = ∫ |𝐷 𝑏 (𝑥 , 𝑦 ) | 𝑑𝑙 (8) Thus, var{D {f }(x)} is periodic over x with period n w By assuminy that: ∆x that ∆x is the sampling step. It is verified the periodicity by the following: 𝑥 𝑥 𝑐𝑜𝑠 Ɵ 𝑛𝑠𝑖 Ɵ , = (9) 𝑦 𝑦 −𝑛𝑠𝑖 Ɵ 𝑐𝑜𝑠 Ɵ { { }( )} 𝑣𝑎𝑟 𝐷 𝑓 𝑥 + 𝜗𝛥 It is possible to represent the Radon transform in = 𝜎 D {𝑤 } − k the following way: { { } } = 𝜎 D {𝑤 } − (k − 𝜗 = 𝑟𝑣𝑎 𝐷 𝑓 (𝑥 ) , , 𝜌 Ɵ(𝑥 ) = 𝐷 {𝑏 (𝑥 , 𝑦 )}. (𝑥 𝑐𝑜𝑠 Ɵ − , , , , 𝑦 𝑛𝑠𝑖 Ɵ, 𝑥 𝑛𝑠𝑖 Ɵ + 𝑦 𝑐𝑜𝑠 Ɵ)𝑑𝑦 (7) (10) Re-sampled signal: From Eq. 4, it is clear that different The Radon transformation is computed at angles θ interpolators will change the structure of the original from 0-179 degrees, in 1 degree increments. Hence, the signal in different ways. The resulting periodic variance output of this is 180 1-D vectors, δθ (θ is the orientation function computed using Eq. 4 for the nearest-neighbor of the X’ axis counterclockwise from the x-axis). The interpolation. Hence, signals interpolated by this corresponding auto covariance sequences of δθ contain a specific strong periodicity, if the investigated signal has interpolator are easily recognized by applying a been re-sampled. The auto covariance sequences of θ derivative operator to them (Hou and Andrews, 1987). It is computed as Eq.11: is continuous but its first derivative is discontinuous. Cubic interpolation is a very frequently used 𝑅 (𝑘 ) = ∑ (ρθ ( i + k) − ρθ )( ρθ ( i) − ρθ ) (11) interpolation technique and has been widely studied. It Ɵ uses a third order interpolation polynomial as the kernel. The project target is to determine the image being Radon transformation: Radon transformation is investigated has undergone affine transformation. applied to find traces of affine transformation. The Hence, we focus only on the strongest periodic patterns Radon transformation computes projections of present in the auto covariance Rδθ. 3 End time Fig. 2: Proposed system performance flowchart Image verification (Phase 1): The image verification Image recognition: The difficulty of recognizing faces phase is performed using detecting traces of image is a classification problem. The proposed technique tampering such as re-sampling or interpolation or both. uses artificial neurons in order to train the classifier. It In this phase, a test/input process involves derivative also involves a dimensionality reduction operators and radon transformation. This process preprocessing of the facial images. In recent times, produces a periodic pattern in the image spectrum if the BPNN system (Vinay et al., 2007) implementation test image is forged or it produces simple impulse make use of fractal encoding method, the fractal codes signals if the test image is still original. were presented as input to the BPNN for identification Meanwhile, It is observed that both image spectra forms could be overlapped that produces a form of image purposes. However, in order to complete image tampering. However, the proposed system specification authentication and recognition in the research work, allows selecting only original test or input images. four stages are introduced. Proposed system structure: The proposed research work deals with an integration of four phases. The initial phase deals with image verification (originality authentication) with neural network pre-processing training and testing phases. Figure 2 shows the flowchart performance of this research work that will be discussed in four stages. 4 End time are input, hidden and output layer. The number of neurons in the hidden layer is determined by experiences and guesswork considering optimal performance. As a final point, production level contains nerve cells that are the number of objects is being considered. The algorithm for effective use of a neural networks and reduces the slope of errors through changing weight and offset continued with impetus. The neural network is trained upon some set of images and tested upon different set of images. In the proposed method, neural Fig. 3: Image vectorization network uses BPNN algorithm for error computation and new weight calculation for each neuron link. It returns the output of each level, extract the mean square error (MSE) and spread it back if it is not close to the target. The response of the neural network is reliant upon weights, biases and transfer functions. The transfer functions make use of in the feed forward BPNN in intermediate, input and output layer. Fig. 4 shows the training graph of neural network while the system will find the original image. Testing stage (Phase 4): The fourth phase is the testing of the neural network. Images for testing are applied to Fig. 4: Neural network training curve the trained neural network along with the already Image-preprocessing (Phase 2): The image trained database images for calculating the percentage preprocessing as second phase is ready to function. In accuracy and error. the pre-processing phase, time effective preprocessing In testing phase, just like in the training phase, is performed in order to make image data best fit for incoming images undergo all the pre-processing stage neural network input. Average filtering is applied and and are made available to the network for evaluation. contrast of the image is enhanced through histogram This test images (extracted face image) is then equalization process. processed using the neural network analytic tool. After Then the image size is reduced in order to make it a number of iterations by the network through each light but efficient and best fit for neural processing image in the database, the error reduces based on the phase. Before moving images into the neural processing gradient descent-learning rule, the set error goal is phase, all images data (test image and training database reached and a matching image to test image is found as images) undergo a process of vectorization that is shown in the results session. conversion of 2-D images into 1-D vectors. This RESULTS AND DISCUSSION conversion is due to neural network requires 1-D vectors for processing and conversion is performed The two database containing training images is using PCA (Esbensen and Geladi, 1987) and 2-D made up of ten images numbered (1-10) each, with the discrete cosine transform. Figure 3 shows the same size, format and dimension. The test images are vectorization of one original image. sampled and cropped with the different position. Figure 5 and 6 show the image spectrum respectively. Neural network training (Phase 3): The third phase is the neural network training. The structure of the proposed neural network is based upon multi-layers that 5 End time Fig. 5: Original image spectrum Fig. 6: Forged image spectrum Fig. 7: Image recognition 6 End time Fig. 8: Neurons in hidden layer vs. execution time Fig. 9: Recognition percentage vs. number of subjects These spectra are resulted from original and forged The major integral part of the proposed analytical tool images respectively. The difference in spectrum is seen is the capability to extract a face from a full image and run as a kind of sinusoidal periodic pattern, which replaces it through any given database for a matching or equivalent or overlaps the original impulse signals or spikes of an face image. Figure 7 illustrates the recognition of input original image. These are the result of obtaining the image while different position is applied. first and second derivative of the auto covariance of the The high-speed performance is determined by the images. The auto covariance is obtained using the radon proper selection of the number of hidden neurons. The transformation through 0-179 degrees. Hence, it is a larger amount of hidden neurons results the faster proved for efficiency of proposed system to detect the network converges. Figure 8 shows the time consuming authenticity of images. 7 End time 304-311. Graz, Austria, April 1-3. Published by required when the hidden layers in neural network Springer-Verlag. DOI: 10.1.1.89.1918 processor are increased. Esbensen K, Geladi P. 1987. Chemometrics and Figure 9 shows the error percentages of proposed Intelligent Laboratory Systems. Proceedings of the analytical tools when the number of imaged are Multivariate Statistical Workshop for Geologists increased. As shown in Fig. 9, it is found that with and Geochemists, Volume 2, Issues 1–3, pp. 37– increasing the number of subjects, performance error 52. DOI: 10.1016/0169-7439(87)80084-9 that is percentage of recognition with utilizing PCA and Gallagher, A.C., 2005. Detection of linear and cubic BPNN approach will decrease slightly. interpolation in JPEG compressed images. In Proceeding of the IEEE Computer. Soc 2nd MATERIALS AND METHODS Canadian Conf. Computer Robot Vision, Washington, DC, pp: 65-72. DOI: The analytical image authentication tool designed 10.1109/CRV.2005.33 based on graphical interfacing computer-aided system. Han, H., J. Jeong and E. Arai, 2011. Virtual out of The design process was to create software model of focus with single image to enhance 3D perception. efficient tool using MATLAB-GUI softwar. The system IEEE Conference on 3DTV The True Vision- was designed simulated for image authentication and Capture, Transmission and Display of 3D Video, classification. The design was emulated the following pp: 1-4 DOI: 10.1109/3DTV.2011.5877188 characteristics; neureal network processing, local binary Hou, H. and H. Andrews, 1978. Cubic splines for image pattern and finaly analytical tools smart system. interpolation and digital filtering. IEEE Trans. Acoust., Speech Signal Process., 26: 508-517. CONCLUSION DOI: 10.1109/TASSP.1978.1163154 Jaeyoung, K. and H. Jun, 2011. Implementation of In this research work, an integration of image image processing and augmented reality programs forgery detection with an image facial recognition for smart mobile device. IEEE Conference on analytical tool using back propagation neural network Strategic Technology (IFOST). 2: 1070-1073. DOI: was proposed. In this project the test image is 10.1109/IFOST.2011.6021205 authenticated before being fed into the recognition tool. Kumar, V.B., B.S. Shreyas and G.C.N.S. Murthy, 2007. Therefore, the result of the first image verification of A Back Propagation Based Face Recognition this project confirms the originality or alteration of an Model, Using 2D Symmetric Gabor Features. IEEE image before it is feed into the rest of the algorithm for Proceeding of The Signal Processing, recognition purpose. It also introduces contrast Communications And Networking, pp: 433-437. stretching of histograms, dual database and adjustable percentage error as a functionality of the proposed DOI: 10.1109/ICSCN.2007.350776 research work. The BPNN for face image recognition is Lu, J., K.N. Plataniotis and A.N. Venetsanopoulos, only highly accurate with small number of subjects and 2003. Regularized Discriminate Analysis For the requires a lowering of the image resolution for a whole Small Sample Size Problem in Face recognition, recognition task as a way of reducing the computational Science Direct Pattern Recognition Letters. Sci. complexity. The project was designed and investigated Direct 24: 3079-3087. DOI: 10.1016/S0167- and it was found that the resolution improved by ±2% 8655(03)00167-3 error rate when number of database images applied is Mehryar, S., K. Martin, K.N. Plataniotis and S. less than 20 images. Stergiopoulos, 2010. Automatic landmark detection for 3D face image processing. IEEE REFERENCES Conference on Evolutionary Computation (CEC), pp: 1-7. DOI: 10.1109/CEC.2010.5586520 Ahonen, T. and A. Hadid, 2006. Face recognition with Ming, H., Q. Zhang and Z. Wang, 2008. Application of local binary patterns: Application on Face Rough Sets to Image Pre-Processing for Face Recognition, Proc. IEEE Transactions on Pattern Detection. IEEE Proc. Of the information and Analysis and Machine Intelligence IEEE Press, 28: automation, pp: 545-548. DOI: 2037-2041. DOI: 10.1109/TPAMI.2006.244 10.1109/ICINFA.2008.4608060 Bolme, D., R. Beveridge, M. Teixeira and B. Draper, Popescu A.C. and H. Farid, 2005. Exposing digital 2003. The CSU Face Identification Evaluation forgeries by detecting traces of re-sampling. IEEE System: Its Purpose, Features and Structure. Trans. Signal Process., 53: 758-767. DOI: International Conference on Vision Systems, pp: 10.1109/ICOSP.2006.345714 8 End time Prasad, S. and K.R. Ramakrishnan, 2006. On re- sampling detection and its application to image tampering. in cedinging of the IEEE Interenational Conference Multimedia Expo., Toronto, ON, Canada, pp: 1325-1328. DOI: 10.1109/ICIEA.2009 Sirovich, L. and M. Kirby, 1987. A low dimensional procedure for the characterization of human faces. J. Optical Society Am., 4:519-524. DOI: 10.1364/JOSAA.4.000519 Turk, M.A. and A.P. Pentland, 1991. Face recognition using eigenfaces. IEEE Conference on Computer Vision and Pattern Recognition, pp: 586-591. DOI: 10.1109/CVPR.1991.139758 Wiskott, L., J.M. Fellous, N. Kuiger and C. von der Malsburg, 1997. Face recognition by elastic bunch graph matching. IEEE Trans. On pattern Analysis and machine intelligence, 19: 775-779. DOI: 10.1109/34.598235 Wu, C.J. and J.S. Huang, 1990. Human face profile recognition by computer. J. Sci. Direct Pattern Recognition, 23: 255-259. DOI: 10.1016/0031- 3203(90)90013-B http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Electrical Engineering and Systems Science arXiv (Cornell University)

Design an Advance computer-aided tool for Image Authentication and Classification

Loading next page...
 
/lp/arxiv-cornell-university/design-an-advance-computer-aided-tool-for-image-authentication-and-aoG4nwmNiE

References (21)

ISSN
1546-9239
eISSN
ARCH-3348
DOI
10.3844/ajassp.2013.696.705
Publisher site
See Article on Publisher Site

Abstract

In/SN Date: 30-05-2012 Time: 1: 30 Design an Advance Computer-Aided Tool for Image Authentication and Classification 1.2 1 Rozita Teymourzadeh, Amirize Alpha Laadi, 2 2 2 Yazan Samir Algnabi, MD Shabul Islam and Masuri Othman Department of Electrical and Electronic, Faculty of Engineering, Technology and Built Environment UCSI University, Jalan Choo Lip Kung, Taman Taynton View, 56000 Cheras, Kuala Lumpur, Malaysia Department of VLSI Design, Institute of Micro Engineering and Nanoelectronics IMEN, Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor, Malaysia Abstract: Problem statement: Over the years, advancements in the fields of digital image processing and artificial intelligence have been applied in solving many real-life problems. This could be seen in facial image recognition for security systems, identity registrations. Hence a bottleneck of identity registration is image processing. Approach: These are carried out in form of image preprocessing, image region extraction by cropping, feature extraction using Principal Component Analysis (PCA) and image compression using Discrete Cosine Transform (DCT). Other processing include filtering and histogram equalization using contrast stretching is performed while enhancing the image as part of the analytical tool. Hence, this research work presents a universal integration image forgery detection analysis tool with image facial recognition using Black Propagation Neural Network (BPNN) processor. The proposed designed tool is a multi-function smart tool with the novel architecture of programmable error goal and light intensity. Furthermore, its advance dual database increases the efficiency for high performance application. Results: With the fact that, the facial image recognition will always, give a matching output or closest possible output image for every input image irrespective of the authenticity, the universal smart GUI tool is proposed and designed to perform image forgery detection with the high accuracy of ±2% error rate. Conclusion: Meanwhile, a novel structure that provides efficient automatic image forgery detection for all input test images for the BPNN recognition is presented. Hence, an input image will be authenticated before being fed into the recognition tool. Key words: Principal Component Analysis (PCA), Discrete Cosine Transform (DCT), Black Propagation Neural Network (BPNN), Local Binary Pattern (LBP) INTRODUCTION 40 numbers with admixtures of Eigen pictures can be characterized a face. Facial image recognition is a bottleneck of image Wu and Haung (1990) took advantage of 24 processing that shows a lot of interrest in past few years measurements in fiducially signs of profile double (Kumar et al., 2007). Move towardes to 2-D face facial image. The proposed system has been reduced by recognition leads to have template-based approach and recognizing performance when more image in the geometric-based approach (Bolme et al., 2003). further database because there are not enough to differentiate complex features are composed of several of the feature features that identify a user. Then, Turk and Penland (1991) introduced the examples. This type of representation leads to an Eigen faces method that achieved a remarkable abstraction from the image pixels. acknowledgement speed, but the speed drastically Sirovich and Kirby (1987) proposed Eigen picture reduced, when against resizing. Wiskott et al. (1997) as economically representation of image in a best used magnitude information and phase information that coordinate system. It was observed that an acceptable is received from Gabor wavelet transformation of the picture of a face could be reconstructed from the face image. For identifying a person, this information is specification of gray levels at 214 pixel locations. compared with all of the information in database However, they showed that in actual construction, only Corresponding Author: Rozita Teymourzadeh, Department of Electrical and Electronic, Faculty of Engineering, Technology and Built Environment, UCSI University, 56000, Cheras Kuala Lumpur Malaysia 1 End time searching for the one who has the same transformation as X΄  T  X, Y Y΄  T  X, Y (1) X Y test image. As earlier stated, the geometric operations that are Gallagher (2005) analyzed a set of images and common with most forgeries are re-scaling, rotation and proposed that image forgery most often includes some skewing. Hence, the importance of detection will be form of geometric transformation. These processes traced wit utilization of affine transformations. were based on a re-sampling and interpolation step. The general equations for affine transformations Later on, Ahonen and Hadid (2006) introduced the are given by: Local Binary Pattern (LBP). This operator was used to X΄  a  a x  a y 0 1 2 measure the texture information of local area of gray (2) Y΄  b  b x  b y 0 1 2 image. The second stage is the interpolation. Here pixel intensity values of the transformed image are assigned The various research work had been completed using low pass interpolation filter. According to the (Ming et al., 2008; Mehryar et al., 2010; Jaeyoung and sampling theory, if the Nyquist criterion is satisfied, the Jun, 2011; Han et al., 2011; Lu et al., 2003) to reduce spectrum F(w) does not overlap in the Fourier domain. the efficiency with variations in illumination, image The original signal f(x) can be reconstructed perfectly size, size of database, percentage error. In addition, from its samples f using the optimal sine interpolation inability to detect altered images for more realistic (Hou and Andrews, 1987). recognition still is challenging task. Combining the derivative theorem with the Hence, this study proposes an efficient passive convolution theorem leads to the conclusion that by evaluation tool for authenticity of input images before convolution of f with derivative kernel D (w), it is k n recognition process stage, furthermore, the adjustment possible to reconstruct the nth derivative of the image of illumination of images using contrast stretching and image histogram equalization techniques will be f(x). The result of interpolated operation of f (x) is performed. In this proposed research work, the quality denoted by Equ. 3; percentage is adjustable in order to achieve high performance (as low as 2% error rate) with a dual 𝑓 (𝑥 ) = f w − k database approach. The test image is reduced to a lower 1-D dimensional vector, which represents D {f }(𝑥 ) = D f w − k = distinguishing features of the test image. The 1-D image vectors are fed into the neural f D {w} − k (3) network, the Euclidean distance of the 1-D image vector is compared to each test database and the closest match is found and outputted. By assuming the constant variance random process, n w then the variance of D {f } which is var {D {f }(x)} n w System design fundamental: The focus of the image as a function of x is given by: authentication stage (Fig. 1) is the presence of the { { } } image re-sampling trace and/or interpolation signal. A 𝑟𝑣𝑎 𝐷 𝑓 (𝑥 ) = 𝑅 {𝑓 }(𝑥 , 𝑥 ) = derivative operator on the variance of the image will 𝜎 D {𝑤 } − k (4) detect these traces. Later, the signal is processed through radon transformation that results in a periodic signal embedded or completely imposed on the image Similarly, the covariance is represented as Equ. 5: spectrum graph (Gallagher, 2005). Hence, interpolated signal and re-sampled signal will be ready to function. 𝑅 {𝑓 }(𝑥 , 𝑥 + 𝜉 ) = 𝜎 D {𝑤 } − k Interpolated signal: There are two main stages in xD {𝑤 } − k (5) geometric transform (Popescu and Farid, 2005; Prasad & Ramakrishnan, 2006). In the first stage, a spatial transformation of the physical re-arrangement of pixels in the image is performed and can be represented by a transformation function T as shown in Equ.1. 2 End time Fig. 1: Proposed process structure Now, by assuming that 𝜗 is an integer, it can be magnitude of Dn{b(x, y)} along specified direction determination by an angle θ. noticed that: The projection is a line integral in a certain direction. This line integral is expressed as: { { }( )} { { }( )} 𝑟𝑣𝑎 𝐷 𝑓 𝑥 = 𝑟𝑣𝑎 𝐷 𝑓 𝑥 + 𝜗𝛥 (6) { } { } 𝜌𝐷 𝑏 (𝑥 , 𝑦 ) = ∫ |𝐷 𝑏 (𝑥 , 𝑦 ) | 𝑑𝑙 (8) Thus, var{D {f }(x)} is periodic over x with period n w By assuminy that: ∆x that ∆x is the sampling step. It is verified the periodicity by the following: 𝑥 𝑥 𝑐𝑜𝑠 Ɵ 𝑛𝑠𝑖 Ɵ , = (9) 𝑦 𝑦 −𝑛𝑠𝑖 Ɵ 𝑐𝑜𝑠 Ɵ { { }( )} 𝑣𝑎𝑟 𝐷 𝑓 𝑥 + 𝜗𝛥 It is possible to represent the Radon transform in = 𝜎 D {𝑤 } − k the following way: { { } } = 𝜎 D {𝑤 } − (k − 𝜗 = 𝑟𝑣𝑎 𝐷 𝑓 (𝑥 ) , , 𝜌 Ɵ(𝑥 ) = 𝐷 {𝑏 (𝑥 , 𝑦 )}. (𝑥 𝑐𝑜𝑠 Ɵ − , , , , 𝑦 𝑛𝑠𝑖 Ɵ, 𝑥 𝑛𝑠𝑖 Ɵ + 𝑦 𝑐𝑜𝑠 Ɵ)𝑑𝑦 (7) (10) Re-sampled signal: From Eq. 4, it is clear that different The Radon transformation is computed at angles θ interpolators will change the structure of the original from 0-179 degrees, in 1 degree increments. Hence, the signal in different ways. The resulting periodic variance output of this is 180 1-D vectors, δθ (θ is the orientation function computed using Eq. 4 for the nearest-neighbor of the X’ axis counterclockwise from the x-axis). The interpolation. Hence, signals interpolated by this corresponding auto covariance sequences of δθ contain a specific strong periodicity, if the investigated signal has interpolator are easily recognized by applying a been re-sampled. The auto covariance sequences of θ derivative operator to them (Hou and Andrews, 1987). It is computed as Eq.11: is continuous but its first derivative is discontinuous. Cubic interpolation is a very frequently used 𝑅 (𝑘 ) = ∑ (ρθ ( i + k) − ρθ )( ρθ ( i) − ρθ ) (11) interpolation technique and has been widely studied. It Ɵ uses a third order interpolation polynomial as the kernel. The project target is to determine the image being Radon transformation: Radon transformation is investigated has undergone affine transformation. applied to find traces of affine transformation. The Hence, we focus only on the strongest periodic patterns Radon transformation computes projections of present in the auto covariance Rδθ. 3 End time Fig. 2: Proposed system performance flowchart Image verification (Phase 1): The image verification Image recognition: The difficulty of recognizing faces phase is performed using detecting traces of image is a classification problem. The proposed technique tampering such as re-sampling or interpolation or both. uses artificial neurons in order to train the classifier. It In this phase, a test/input process involves derivative also involves a dimensionality reduction operators and radon transformation. This process preprocessing of the facial images. In recent times, produces a periodic pattern in the image spectrum if the BPNN system (Vinay et al., 2007) implementation test image is forged or it produces simple impulse make use of fractal encoding method, the fractal codes signals if the test image is still original. were presented as input to the BPNN for identification Meanwhile, It is observed that both image spectra forms could be overlapped that produces a form of image purposes. However, in order to complete image tampering. However, the proposed system specification authentication and recognition in the research work, allows selecting only original test or input images. four stages are introduced. Proposed system structure: The proposed research work deals with an integration of four phases. The initial phase deals with image verification (originality authentication) with neural network pre-processing training and testing phases. Figure 2 shows the flowchart performance of this research work that will be discussed in four stages. 4 End time are input, hidden and output layer. The number of neurons in the hidden layer is determined by experiences and guesswork considering optimal performance. As a final point, production level contains nerve cells that are the number of objects is being considered. The algorithm for effective use of a neural networks and reduces the slope of errors through changing weight and offset continued with impetus. The neural network is trained upon some set of images and tested upon different set of images. In the proposed method, neural Fig. 3: Image vectorization network uses BPNN algorithm for error computation and new weight calculation for each neuron link. It returns the output of each level, extract the mean square error (MSE) and spread it back if it is not close to the target. The response of the neural network is reliant upon weights, biases and transfer functions. The transfer functions make use of in the feed forward BPNN in intermediate, input and output layer. Fig. 4 shows the training graph of neural network while the system will find the original image. Testing stage (Phase 4): The fourth phase is the testing of the neural network. Images for testing are applied to Fig. 4: Neural network training curve the trained neural network along with the already Image-preprocessing (Phase 2): The image trained database images for calculating the percentage preprocessing as second phase is ready to function. In accuracy and error. the pre-processing phase, time effective preprocessing In testing phase, just like in the training phase, is performed in order to make image data best fit for incoming images undergo all the pre-processing stage neural network input. Average filtering is applied and and are made available to the network for evaluation. contrast of the image is enhanced through histogram This test images (extracted face image) is then equalization process. processed using the neural network analytic tool. After Then the image size is reduced in order to make it a number of iterations by the network through each light but efficient and best fit for neural processing image in the database, the error reduces based on the phase. Before moving images into the neural processing gradient descent-learning rule, the set error goal is phase, all images data (test image and training database reached and a matching image to test image is found as images) undergo a process of vectorization that is shown in the results session. conversion of 2-D images into 1-D vectors. This RESULTS AND DISCUSSION conversion is due to neural network requires 1-D vectors for processing and conversion is performed The two database containing training images is using PCA (Esbensen and Geladi, 1987) and 2-D made up of ten images numbered (1-10) each, with the discrete cosine transform. Figure 3 shows the same size, format and dimension. The test images are vectorization of one original image. sampled and cropped with the different position. Figure 5 and 6 show the image spectrum respectively. Neural network training (Phase 3): The third phase is the neural network training. The structure of the proposed neural network is based upon multi-layers that 5 End time Fig. 5: Original image spectrum Fig. 6: Forged image spectrum Fig. 7: Image recognition 6 End time Fig. 8: Neurons in hidden layer vs. execution time Fig. 9: Recognition percentage vs. number of subjects These spectra are resulted from original and forged The major integral part of the proposed analytical tool images respectively. The difference in spectrum is seen is the capability to extract a face from a full image and run as a kind of sinusoidal periodic pattern, which replaces it through any given database for a matching or equivalent or overlaps the original impulse signals or spikes of an face image. Figure 7 illustrates the recognition of input original image. These are the result of obtaining the image while different position is applied. first and second derivative of the auto covariance of the The high-speed performance is determined by the images. The auto covariance is obtained using the radon proper selection of the number of hidden neurons. The transformation through 0-179 degrees. Hence, it is a larger amount of hidden neurons results the faster proved for efficiency of proposed system to detect the network converges. Figure 8 shows the time consuming authenticity of images. 7 End time 304-311. Graz, Austria, April 1-3. Published by required when the hidden layers in neural network Springer-Verlag. DOI: 10.1.1.89.1918 processor are increased. Esbensen K, Geladi P. 1987. Chemometrics and Figure 9 shows the error percentages of proposed Intelligent Laboratory Systems. Proceedings of the analytical tools when the number of imaged are Multivariate Statistical Workshop for Geologists increased. As shown in Fig. 9, it is found that with and Geochemists, Volume 2, Issues 1–3, pp. 37– increasing the number of subjects, performance error 52. DOI: 10.1016/0169-7439(87)80084-9 that is percentage of recognition with utilizing PCA and Gallagher, A.C., 2005. Detection of linear and cubic BPNN approach will decrease slightly. interpolation in JPEG compressed images. In Proceeding of the IEEE Computer. Soc 2nd MATERIALS AND METHODS Canadian Conf. Computer Robot Vision, Washington, DC, pp: 65-72. DOI: The analytical image authentication tool designed 10.1109/CRV.2005.33 based on graphical interfacing computer-aided system. Han, H., J. Jeong and E. Arai, 2011. Virtual out of The design process was to create software model of focus with single image to enhance 3D perception. efficient tool using MATLAB-GUI softwar. The system IEEE Conference on 3DTV The True Vision- was designed simulated for image authentication and Capture, Transmission and Display of 3D Video, classification. The design was emulated the following pp: 1-4 DOI: 10.1109/3DTV.2011.5877188 characteristics; neureal network processing, local binary Hou, H. and H. Andrews, 1978. Cubic splines for image pattern and finaly analytical tools smart system. interpolation and digital filtering. IEEE Trans. Acoust., Speech Signal Process., 26: 508-517. CONCLUSION DOI: 10.1109/TASSP.1978.1163154 Jaeyoung, K. and H. Jun, 2011. Implementation of In this research work, an integration of image image processing and augmented reality programs forgery detection with an image facial recognition for smart mobile device. IEEE Conference on analytical tool using back propagation neural network Strategic Technology (IFOST). 2: 1070-1073. DOI: was proposed. In this project the test image is 10.1109/IFOST.2011.6021205 authenticated before being fed into the recognition tool. Kumar, V.B., B.S. Shreyas and G.C.N.S. Murthy, 2007. Therefore, the result of the first image verification of A Back Propagation Based Face Recognition this project confirms the originality or alteration of an Model, Using 2D Symmetric Gabor Features. IEEE image before it is feed into the rest of the algorithm for Proceeding of The Signal Processing, recognition purpose. It also introduces contrast Communications And Networking, pp: 433-437. stretching of histograms, dual database and adjustable percentage error as a functionality of the proposed DOI: 10.1109/ICSCN.2007.350776 research work. The BPNN for face image recognition is Lu, J., K.N. Plataniotis and A.N. Venetsanopoulos, only highly accurate with small number of subjects and 2003. Regularized Discriminate Analysis For the requires a lowering of the image resolution for a whole Small Sample Size Problem in Face recognition, recognition task as a way of reducing the computational Science Direct Pattern Recognition Letters. Sci. complexity. The project was designed and investigated Direct 24: 3079-3087. DOI: 10.1016/S0167- and it was found that the resolution improved by ±2% 8655(03)00167-3 error rate when number of database images applied is Mehryar, S., K. Martin, K.N. Plataniotis and S. less than 20 images. Stergiopoulos, 2010. Automatic landmark detection for 3D face image processing. IEEE REFERENCES Conference on Evolutionary Computation (CEC), pp: 1-7. DOI: 10.1109/CEC.2010.5586520 Ahonen, T. and A. Hadid, 2006. Face recognition with Ming, H., Q. Zhang and Z. Wang, 2008. Application of local binary patterns: Application on Face Rough Sets to Image Pre-Processing for Face Recognition, Proc. IEEE Transactions on Pattern Detection. IEEE Proc. Of the information and Analysis and Machine Intelligence IEEE Press, 28: automation, pp: 545-548. DOI: 2037-2041. DOI: 10.1109/TPAMI.2006.244 10.1109/ICINFA.2008.4608060 Bolme, D., R. Beveridge, M. Teixeira and B. Draper, Popescu A.C. and H. Farid, 2005. Exposing digital 2003. The CSU Face Identification Evaluation forgeries by detecting traces of re-sampling. IEEE System: Its Purpose, Features and Structure. Trans. Signal Process., 53: 758-767. DOI: International Conference on Vision Systems, pp: 10.1109/ICOSP.2006.345714 8 End time Prasad, S. and K.R. Ramakrishnan, 2006. On re- sampling detection and its application to image tampering. in cedinging of the IEEE Interenational Conference Multimedia Expo., Toronto, ON, Canada, pp: 1325-1328. DOI: 10.1109/ICIEA.2009 Sirovich, L. and M. Kirby, 1987. A low dimensional procedure for the characterization of human faces. J. Optical Society Am., 4:519-524. DOI: 10.1364/JOSAA.4.000519 Turk, M.A. and A.P. Pentland, 1991. Face recognition using eigenfaces. IEEE Conference on Computer Vision and Pattern Recognition, pp: 586-591. DOI: 10.1109/CVPR.1991.139758 Wiskott, L., J.M. Fellous, N. Kuiger and C. von der Malsburg, 1997. Face recognition by elastic bunch graph matching. IEEE Trans. On pattern Analysis and machine intelligence, 19: 775-779. DOI: 10.1109/34.598235 Wu, C.J. and J.S. Huang, 1990. Human face profile recognition by computer. J. Sci. Direct Pattern Recognition, 23: 255-259. DOI: 10.1016/0031- 3203(90)90013-B

Journal

Electrical Engineering and Systems SciencearXiv (Cornell University)

Published: Jun 16, 2018

There are no references for this article.