Unfamiliar Sides, Video, Image Enhancement in Face Recognition

International Journal of Hybrid Information Technology Vol. 9, No.11 (2016), pp. 255-266 http://dx.doi.org/10.14257/ijhit.2016.9.11.22 Unfamiliar Sides, Video, Image Enhancement in Face Recognition 1 Ranbeer Tyagi and 2Geetam Singh Tomar Deptt. of EC UTU, Dehradun, India THDC-IHET, Tehri-Garhwal (U.K.),India [email protected] [email protected] O m nl ad in e eV by e th rsio is n fil O e is nly IL . LE G AL . Abstract The majority of image-processing techniques interact treating the impression as being a two dimensional signal and applying normal signal-processing techniques to it. Within this document we offer a new standard approach to have the hidden restrictions which aren't revealed by the Sobel together Canny filters together in face-recognition environment. We've also showed the effect of the contrast and threshold around the photos. In the same manner, we have shown our techniques about the video. Thus, many beautiful contrast and the hidden data are shared. Keywords:Image processing, Enhancement, Edge Detection, Segmentation, Video Image-Processing, Face Recognition 1. Introduction ok In image handling we're dealing with deterministic and stochastic representations of pictures that increase the quality of pictures by eliminating deterioration offered on graphic. Along the way of the image recovery we try to restore an image from changed one so that it is as close that you can for the initial photograph. Some degradation contains arbitrary sound, interference, geometrical distortions, loss of contrast, blurring effects, etc. Modern electronic technology has made it possible to manipulate multiple- dimensional signals with programs that range from easy electronic circuits to advanced parallel computers. The aim of this treatment can be divided in to three classes [1]:  Image in  image out ; Image Processing;  Image in  Picture in measurements out; Image Analysis;  photograph in  picture in high-level explanation out; Picture Comprehension; Image Segmentation Bo Graphic segmentation refers to the major step up image processing where the inputs are images and, results are the capabilities extracted from those photos. Segmentation breaks impression into its component parts or objects [3]. The amount to which segmentation is carried out depends upon the issue being fixed i.e. segmentation must halt if the items of curiosity about a software have already been separated. Image segmentation describes the decomposition of a landscape into its parts. Like in the automatic examination of automated units, curiosity is based on the inspecting photos of the products together with the goal of deciding the presence or absence of unique anomalies, such as missing elements or damaged association paths. There's no point in carrying segmentation past the level of aspect required to determine those things. Segmentation of nontrivial images is one of many hardest responsibilities in image processing. Segmentation precision decides the eventual achievement or malfunction of ISSN: 1738-9968 IJHIT Copyright © 2016 SERSC International Journal of Hybrid Information Technology Vol. 9, No.11 (2016) computerized examination methods. Because of this extensive attention is taken to improve the likelihood of tough segmentation. In certain conditions such as industrial inspection programs, atleast some measure of control over the environment is possible occasionally. In others, as in remote-sensing, user control over image order is limited mainly to the choice of image sensors. Edge Detection O m nl ad in e eV by e th rsio is n fil O e is nly IL . LE G AL . Sides are freely thought as pixel depth discontinuities within an image. Edge detection is just a subjective task. It is an easy task to find these clear ends, or those with superior S/D ratio. Border detector is tailored to benefit from the site knowledge [8], [9]. For example, a "straightedge" alarm may be very effective in locating many structures and objects including golf courts in an aerial image. Ends characterize limits and are consequently a problem of elementary importance in image processing. There are typical prices for your various guidelines encountered in digital image processing. These values can be brought on by video expectations, by algorithmic specifications, or from the need to preserve digital circuitry straightforward. Factors Affecting the Selection of Edge Detector Side inclination: The geometry of the agent determines a characteristic way in which it's many delicate to tips. Providers can be optimized to find horizontal, vertical, or straight edges. Disturbance setting: Border discovery is hard in noisy images, since both noise and the ends contain high-frequency content. Endeavors to cut back the noise end in confused and distorted tips [4]. Operators utilized on loud photographs are typically larger in breadth, for them to average enough knowledge to discount local noisy pixels. This leads to less exact localization of the discovered edges. Edge structure: Not all ends require a step-change in strength. Effects such as refraction or inadequate concentration can result in things with limitations identified by way of a slow change in power [10]. The owner has to be selected to become tuned in to such a gradual change in these scenarios. Newer wavelet-based tactics truly define the character of the move for each side as a way to distinguish. Image Enhancement ok The basic purpose of Photograph improvement is the fact that the observed image have apparent trait that was unavailable for your original graphic. Various methods including intensity transformation histogram equalization, homomorphic filter, and have been suggested to boost photographs deteriorated by irregular light. These methods typically boost an insight graphic by reducing its dynamic-range and or raising its contrast [2], [5]. Bo Face Recognition Nowdays has been the fantastic significance of Picture Repair and face-recognition. The extremely research location has been face recognition in computer-vision in support of the final mix of ages. Face-recognition strategies are often employed for security areas apart from are slowly more finding utilized in numerous various uses [6]. It could determine a complete firm in an electronic photograph by reviewing and evaluating types that's a kind of biometric software goal. For instance, employs facial-recognition to tell apart among people within the The Kinetic movement gaming system. Experienceidentification has been analyzed carefully; nonetheless, real-world face-identification nevertheless remains a hard job. [31] Lei et-al. Research discriminant minimal aesthetic search discussed discriminant experience descriptors (DED) in an info - powerful model in its placement of the handcrafted approach which is effectiveness on together 256 Copyright © 2016 SERSC International Journal of Hybrid Information Technology Vol. 9, No.11 (2016) Bo ok O m nl ad in e eV by e th rsio is n fil O e is nly IL . LE G AL . harmonized experience recognition and varied face recognition. Additional to business in acceptance wherever substantial-dimensional skin persistence aren't chosen, [32] Chen ETAL. the hightech is realized in experientially generate noticeable that in the shape of high-dimensional LBP capability descriptors. To aim the pretense variations, existinginvariant face-recognition is attained by with Markov random meadow stay snapshot harmonizing [33]. Authors suggest a newest limited face-recognition want to identify people of awareness by their fractional activities. Eventually, the likeness of two faces is rehabilitated while the detachment among these two allied characteristic jobs [34]. This establishes the IARPA Bench Mark A (IJB-A), an overtly accessible strategy inside the untamed dataset maintaining 500 themes through truly on a somewhat spot expertise photographs. Investigators might sign up for the offering record in support of info on such chance discharges [35]. An image-correct legion normalization program, on polynomial damage, is ready to increase the resilience of expertise harmonizing below tough instances. This file mostly supplied an improved taking of discriminative legion performance joining. In variation, with “wild”, uncontrolled cohort cases allow reaching the great exhibition [36]. The suggested a purpose of misinformation construction using the number of SORTAE to realize present invariant face recognition [37]. This history available individual effectiveness prepared unconstrained nonetheless-to-still and video-to-video knowledge harmonizing conditions. They applied the facial skin archive of two kinds, LFW and YTF [38]. The minimal twin descriptors considered at this time are minimal twin layout, Nearby level quantization, and Binarized arithmetical picture skin persistence [39]. Important computer data-collection includes, an extensive assortment of troubles examining occlusions, hard poses, and modest assertion and out-of-middle activities, the necessity of experience places as oblique regions and together grayscale and shade photographs [40]. A guide expertise data set that will create simple study within the difficulty of forward to say expertise proof „in the wild‟. These records set may be called from your Celebrities Infront-Account data-set. We assessed the recital higher than several different calculations employing a limited procedure and explained how they humiliate by Front-Entrance to Front-Consideration [41]. Writer offers the techniques for unconstrained experience verification affiliated on cavernous convolutional skin consistency and determine it about the recently boundless IARPA bench mark A (IJB-A) dataset along with concerning the conventional Designated Experience in the Wild dataset. We study the effectiveness of the DCNN approach on currently unconfined complicated knowledge evidence info, IARPA Common A, which maintains encounters through whole present, light, and additional challenging situations [42]. The Designated Encounters in the Open document hasbeen completely used [43]. This investigation heart of awareness on growing a face-recognition plan stick to Main Part Evaluation and Self-Organizing Routes unverified instruction formula [44].We have novice a book statement, Designated Activities in the Open, whose key goals are; 1) offer a massive file of genuine globe knowledge photographs created for the hidden two of the sort similar trouble of knowledge identification, 2) Balanced so as interested in the acceptance-setting-reputation channel, and 3) Permit careful and easy variance of face recognition computations [45]. An indistinct picture could be assessed like a dilemma reason behind a sharpened image along with a sort significant element. Consequently in categorize to retrieve the fast photograph we need to separate the impression addicted to its smear crucial portion and fast photograph. Besides this the problem at this time will be the watch of the cloud kernel [46]. The principal purpose of enhancement will be to approach a picture so our consequence is more desirable as opposed to initial picture for a distinct application. In mainly of these crimes, inside the conventional to make use of organize techniques the illegals were intriguing advantageous asset of a primary issue: the firms do not funding right of convenience by "who we're", except by "what we've", such as for Copyright © 2016 SERSC 257 International Journal of Hybrid Information Technology Vol. 9, No.11 (2016) instance for example identification permits, strategies, balances, BANNER information‟s. No of these belongings are now essential people. Lately, capability switched accessible permitting confirmation of "proper" person character. The region of “biometrics” are called this type of executive. Enveloped in the quantity of biometric identification methods, the bodily practices (fingerprint, experience, genetics) are extra continuous than practices in-efficiency group (keystroke, voice print). Face recognition will be the unique of the several biometric methods to be able to obtain the faculties of together elevated accuracy and small intrusiveness. 2. Problem Formulation O m nl ad in e eV by e th rsio is n fil O e is nly IL . LE G AL . To get the edges we studied slope owner which can be predicated on first-order and second order kind [25]. The initial derivative agent uses some standard attributes like; the initial derivative of the grey amount is unfavorable in the top edge of the transition, positive in the trailing side, and zero within the aspects of continual dull levels. For the twodimensional event we have the outside direction, the straight direction, or an arbitrary direction which may be considered as a mix of the two. The next derivative operator meets the fundamental qualities like [18]; the second kind is negative for that lighting area of the edge, constructive for your dark aspect of the edge, and zero for pixels laying specifically on tips. Centered on this technique distinct owner suggested like: Robert, Prewitt, Sobel, and Laplacian edge detector. A few other approach like FIREWOOD (Laplacian of Gaussian), Canny edge detector and zerocrossing side detector. But these all aren't able to recognize the hidden and weak tips in any graphic therefore planned protocol is a possibility to discover the way to the invisible and weak ends in stationary picture together with shifting pictures (videos) [16]. Within the same fashion proposed protocol is useful to identification the objects that aren't in a position to see due to high-brightness and darkness. Objectives  To have the hidden and weakened sides in an image  To obtain the hidden ends in transferring graphic (video)  To boost the high brightness and night watch of an image and realize the item which are not able to imagine.  To boost the night view of the transferring picture (video). 3. Various Type of Edge Detector ok The edge detection methods collected into two groups: Gradient edge detection and Laplacian edge discovery. Bo 3.1. Gradient Edge Detection Within the traditional edge detector, the incline of picture is computed using first order change [7] [17]. If the slope is above the threshold, there is an object while in the graphic. As regarding to picture g(l, m), the incline of point (l, m) means follows: ] | ( ) [ (1) | The weight of the vector is ( Also its way as ( 258 ) ) ( ⁄ [ ) ] ⁄ (2) (3) Copyright © 2016 SERSC International Journal of Hybrid Information Technology Vol. 9, No.11 (2016) Where and would be the incline in l and m way. Gradient of each pixel of the image is determined utilising the above three equations [24]. In reality, little spot structure convolution is employed to process the photograph. Incline operators include John, Prewitt and Sobel operator. 3.2. Laplacian Edge Detector It is helpful in this case to contemplate utilizing the Laplace function. The Laplacian process searches for zero crossings in the next derivative of the graphic to locate tips [22], [30]. An edge gets the one-dimensional shape of a ramp and calculating the kind of the image could emphasize its location. O m nl ad in e eV by e th rsio is n fil O e is nly IL . LE G AL . (4) ( ( ) ( ) ( ) ( ) ) ( ) ( ( [ ) (5) ( ) (6) ) ( ) ( ) (7) ] 4. Image Enhancement Bo ok Graphic improvement strategy continues to be commonly used in several programs where the subjective quality of picture is essential [28]. The objective of graphic improvement is dependent around the program situations. Comparison is an important aspect in any individual evaluation of image-quality, it can be preventing software for documenting and presenting data collected during exam [11]. Numerous tactics such as depth modification histogram equalization, homomorphic filtering, and have been suggested to enhance photographs deteriorated by abnormal illumination. These processes generally enrich an insight graphic by decreasing its dynamic range and-or raising its comparison [12]. Picture advancement is the means of generating pictures more useful. The causes for achieving this include:  Highlighting fascinating depth in images  Removing noise from pictures  Building images more aesthetically appealing Photograph enlargement will be the firststep in image processing. The goal of photograph enhancement would be to enhance the interpretability or perception of on in photographs for human people, or even to supply `better' insight for additional computerized image processing methods [13], [27]. Picture enlargement approaches fall under two broad groups: spatial domain methods and frequency-domain approaches. The word spatial site identifies the image plane itself, and approaches in this classification are derived from primary adjustment of pixels within an impression. Frequency domain control strategies are based on altering the Fourier transform of a graphic. Copyright © 2016 SERSC 259 International Journal of Hybrid Information Technology Vol. 9, No.11 (2016) 4.1. Energy Legislation Modification The productivity image of the energy law transformation relates to its input image. [ ( )] ( ) (8) c and γ are frequent.The importance of γ decides the amount to that the intensity range increases. In an electric law change, each pixel of the original photograph is lifted to certain exponent value. By selecting exponent price correctly you can boost sometimes large or low luminance price [14], [19], [29]. 4.2. SIGMOID Function O m nl ad in e eV by e th rsio is n fil O e is nly IL . LE G AL . Sigmoid function can be a non-linear function. The title sigmoid gets from the proven fact that the function is “S” formed statisticians contact this function the logistic function using e(m) for that feedback and with g being a get [11], [21]. The sigmoid function is distributed by: (9) g is gain which handles the specific distinctionIt‟s having often non-negative and nonconstructive first derivative or just on deflection point. It maps full-range on-scene luminance also this purpose guarantees that no graphic place is condensed and distinction maybe highly condensed. 4.3. GAMMA Correction Gamma correction, gamma nonlinearity, gamma selection, or usually just gamma, could be the brand of the nonlinear operation used-to code and decode luminance prices in video or however graphic devices [23], [26]. In image processing symbol γ signifies the numerical parameter which describes the nonlinearity of strength duplication. The procedure of pre-research for that nonlinearity by research indication from an power value is known as gamma correction [15], [20]. 5. Recommended Formulas (a) Algorithm for Edge Detection [ ] Figure 1. Planned User Bo ok If we alter the Laplacian operator assistance to the evaluation, the following matrix is possible. That is shown in Figure 1 and the steps of algorithm as follows: 1. 2. 3. 4. 5. 6. 7. 260 Take unique image and convolve with the proposed driver Result of move 1 is handed although canny edge detector Today applying histogram equalization is employed on unique graphic Consequence of move 3 is handed Canny edge detector Consequence of stage 3 is approved Zerocrossing edge detector Results of stage 4 and 5 is combined with OR operation Add results of step 6 and 2. Copyright © 2016 SERSC International Journal of Hybrid Information Technology Vol. 9, No.11 (2016) (b) Simulation Model for Video Edge Detection Using proposed driver analyze the video edge discovery together with the support of Automobile thresholding block and fif.2 shows the simulation model for video edge detection. Image Video Viewer Video Viewer cat _video .bin Convert Image to double Y' I1 2-D CONV I2 Image Data Type Conversion Read Binary File 2-D Convolution I Autothreshold BW Video Viewer O m nl ad in e eV by e th rsio is n fil O e is nly IL . LE G AL . Image Autothreshold Video Viewer 2 -C- Constant I Sobel Edge Image Edge Detection Video Viewer Video Viewer 3 Figure 2. Simulink Model for Video Edge Detection (c) Algorithm for Image Development The steps of Image Enhancement Algorithm as follows: 1. Consider initial graphic and pass-through strength legislation change using [ ( ) eq. ( )] . 2. Withhold the minimal value of phase I and split with maximum of results of phase 1. . 3. Consequence of move 3 is approved through sigmoid function using eq. 4. Take the minimum price of stage 3 and split with maximum of consequence of action 3. 5. Complete caused by phase 4 through gamma correction using eq. ( ) [ ( )] . (d) Simulation Model for Video Enhancement Bo ok In this section Figure 3 represents the simulation for video enhancement. Copyright © 2016 SERSC 261 International Journal of Hybrid Information Technology Vol. 9, No.11 (2016) 0.8 1 cutoff Constant 2 vipmen .avi V: 120 x160 ,I 30 fps From Multimedia File u power y u sigmoid y Subtract e Embedded MATLAB Function Embedded MATLAB Function 1 Product u u Math Function u Add pre y gamma y z Divide Embedded Embedded MATLAB Function 2 MATLAB Function 3 Image Video Viewer Video Viewer .9 gain Video Image Viewer 1 Constant 1 1 Constant 3 Divide 1 Video Viewer 1 9 O m nl ad in e eV by e th rsio is n fil O e is nly IL . LE G AL . gamma Figure 3. Simulink Model for Video Enhancement 6. Results and Comparision (a)Result of Video Edge Detection Figure 4. Original Video Figure 5. Sobel Edge Video Figure 6. Advantage Video of Proposed Product Figure 2 displays the design for evaluating the side movie using Sobel edge sensor and proposed driver. Figure 4 demonstrates the initial video. Figure shows video using Sobel edge sensor. Figure 6 displays video using recommended operator. Bo ok (b) Result of Image Enhancement Figure 7. Initial Photograph Figure 8. Figure 9. Strength Comparison Image Impression The proposed formula for the graphic enhancement as demonstrated in section V (C) for the affirmation of our benefits we have considered the following impression as demonstrated in Figure 7. Figure 8 shows the impression when it's passed through the planned algorithm when parameter gamma is set to value 2, gain is about to benefit 3, cutoff is ready to benefit 0.8 and 2nd gamma is ready to price 0.6. This demonstrates the far end target plainly which can be not observable in original impression due to high- 262 Copyright © 2016 SERSC International Journal of Hybrid Information Technology Vol. 9, No.11 (2016) brightness. Figure 9 shows the strength image when it is approved through the proposed algorithm when parameter gamma is about to worth 2, gain is ready to benefit 5, cutoff is set to worth 0.4 and 2nd gamma is set to price 9. Within this photograph the dim part of the first impression is clearly visible. This implies the result photograph is dependent upon different benefit of guidelines. Here gain controls the real distinction. Cutoff is set for the value in a manner the grey benefit regarding which comparison is more than before or decreased. (c) Results for Video Enhancement O m nl ad in e eV by e th rsio is n fil O e is nly IL . LE G AL . The proposed model for our video enlargement as proven in section V (d) for proof of our results we have regarded these movies as revealed in Figure 10. Figure 11 shows the superior video when initial movie handed through proposed type. Figure 10. Unique Movie ok Figure 11. Enhanced Movie 7. Conclusion Bo In this report we've beautifully compared the available link between the Sobel as well as the canny filter response in face recognition. Canny edge detector is used which is one of many most powerful edge detector as the edge points decided, give rise to ridges in the slope size graphic. The canny algorithm subsequently tracks at the very top of these ridges and sets to zero all pixels that are not really around the ridgetop to be able to offer slim range as the output, an activity known as non optimum reduction. Furthermore as a result of usage of two thresholds, unlike other edge detection it's able to discover little power modifications in a impression as ends. Within our method we unearthed that the concealed limits can certainly and properly are scored, those were unseen in Sobel as well as in Canny responses along with boundary of materials is outlined. Our image enhancement and video experimental results indicate that the recommended picture Copyright © 2016 SERSC 263 International Journal of Hybrid Information Technology Vol. 9, No.11 (2016) enlargement approach could dramatically increase the functionality of face detection protocol due to its solid power to improve the graphic awareness. References [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] Bo [22] O m nl ad in e eV by e th rsio is n fil O e is nly IL . LE G AL . [3] I. T. Young Jan, J. Gerbrands and L. J. Van Vliet, “Fundamental of Image Processing”, (1995). D. H. Choi, I. H. Jang, M. H. Kim and N. C. Kim, “Color image enhancement using single-scale retinex based on an improved image formation model”, 16th European Signal Processing Conference (EUSIPCO 2008), Lausanne, Switzerland, (2008). L. Fallah Araghi and R.A. Mohammad, “An Implementation Image Edge and Feature Detection Using Neural Network” Proceedings of the International MultiConference of Engineers and Computer Scientists 2009, Vol I IMECS 2009, (2009). M. H. Asmare, V. S. Asirvadam and L. Iznita, “Color Space Selection for Color Image Enhancement Applications”, International Conference on Signal Acquisition and Processing, (2009). C. Sekhar Panda and S. Patnaik, “Filtering Corrupted Image and Edge Detection in Restored Grayscale Image Using Derivative”, International, Journal of Image Processing, (IJIP), vol. 3, issue 3, (2009). K. S. Sim, L. W. Thong, M. A. Lai and C. P. Tso, “Enhancement of Optical Images using Hybrid Edge Detection Technique”, Conference on Innovative Technologies in Intelligent Systems and Industrial Applications (CITISIA 2009) Monash University, Sunway campus, Malaysia, (2009). M. Roushdy, “Comparative Study of Edge Detection Algorithms Applying on the Grayscale Noisy Image Using Morphological Filter”, GVIP Journal, vol. 6, issue 4, (2006). S. Mahbub Hafiz, Md. N. Hasan and Md. M. Islam, “An Efficient Scanning Based Learning Free Algorithm for Face Detection”, International Conference on Infonnatics, Electronics & Vision (ICIEV), (2012), pp. 52-56. A. Punnappurath, A. Narayanan Rajagopalan, S. Taheri and R. Chellappa, “Face Recognition across Non-Uniform Motion Blur, Illumination and Pose”, IEEE Transactions on image processing, vol. 24, no. 7, (2015). R.C. Gonzalez and R. E. Woods, “Digital image processing”, third edition, (2008). N. Hassan and N. Akamatsu, “A new approach for contrast enhancement using sigmoid function”, Inernational arab journal of information technology, vol. 1, no. 2, (2004) M. Kafai, L. An and B. Bhanu, “Reference Face Graph for Face Recognition”, IEEE Transactions on information forensics and security, vol. 9, issue 12, (2014), pp. 2132 - 2143. K. Rajadnya and K.T. Talele, “Image Enhancement Techniques Pixel Operation in Spatial Domain”, Proceedings of SPIT-IEEE Colloquium and International Conference, Mumbai, India. D. M. Scott and H. Mccann, “Process imaging for automatic control”, (2005). K. Plataniotis and A.N. Venetsanopoulos, “color image processing and application”, (2000). F. Samopa and A. Asano, “Hybrid image thresholding method using edge detection”, IJCSNS International Journal of Computer Science and Network Security, vol. 9, no.4, (2009). Y. Ya and H. Ju, “A sub - pixel edge detection method based on canny operator”, Sixth International Conference on Fuzzy Systems and Knowledge Discovery, (2009). S. Rachmawati Yahya, S. N. H. Sheikh Abdullah, K. Omar, M. S. Zakaria and C. Y. Liong, “Review On Image Enhancement Methods Of Old Manuscript With The Damaged Background” International Conference on Electrical Engineering and Informatics, (2009). D. Zhao, “Image Processing and Feature Extraction”, Electrical and Computer Engineering, (2005). Tao, Li, M.-J. Seow, V. K. Asari, J. T. Astola, K. O. Egiazarian, N. M. Nasrabadi and S. A. Rizvi, “Image Processing Algorithms and Systems Neural Networks and Machine Learning”, (2006). K.S. Sim, “Enhancement of optical images using hybrid edge detection technique”, 2009 Innovative Technologies in Intelligent Systems and Industrial Applications, (2009). X. Huang, R. Netravali, H. Man, V. Lawrence, S. S. Agaian, A. P. Gotchev, J. Recker and G. Wang, “Image Processing Algorithms and Systems X and Parallel Processing for Imaging Applications II”, (2012). S. N. Dhage, A.A. Mishra and R. Patil, “International Conference on Graphic and Image Processing (ICGIP 2011)”, (2011). Joshi, S. Ram and R. Koju, “Study and comparison of edge detection algorithms”, 2012 Third Asian Himalayas International Conference on Internet, (2012). Marshnil, N. Ameena and M. C. Binish, “Image encryption based on diffusion process and multiple chaotic maps”, 2014 International Conference on Power Signals Control and Computations (EPSCICON), (2014). C. Jocelyn, A. K. C. Allen, T. Miller, N. D. Kalka and A.K. Jain, “Grouper: Optimizing Crowdsourced Face Annotations”, To appear in the CVPR Workshop on Biometrics, (2016). L. Best-Rowden, H. Han, C. Otto, B. Klare and A. K. Jain, “Unconstrained Face Recognition: Identifying a Person of Interest from a Media Collection”, To Appear in IEEE Transactions on Information Forensics and Security, (2014). ok [1] [2] [23] [24] [25] [26] [27] 264 Copyright © 2016 SERSC International Journal of Hybrid Information Technology Vol. 9, No.11 (2016) Bo ok O m nl ad in e eV by e th rsio is n fil O e is nly IL . LE G AL . [28] L. Best-Rowden and A. K. Jain, “Longitudinal Study of Automatic Face Recognition”, MSU Technical Report MSU-CSE-15-20, (2015). [29] C. Jocelyn, A. K.C. Allen, T. Miller, N. D. Kalka and A. K. Jain, “A Comparison of Humanand Automated Face Verification Accuracy on Unconstrained Image Sets”, To appear in the CVPR Workshop on Biometrics, (2016). [30] J. Cheney, B. Klein, A. K. Jain and B. F. Klare, “Unconstrained Face Detection: State of the Art Baseline and Challenges”, the 8th IAPR International Conference on Biometrics (ICB), Phuket, Thailand, (2015). [31] Z. Lei, M. Pietikainen and S. Li, “Learning discriminant face descriptor”, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 36, no. 2, (2014), pp. 289–302. [32] D. Chen, X. Cao, F. Wen and J. Sun, “Blessing of dimensionality: Highdimensional feature and its efficient compression for face verification”, in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), (2013), pp. 3025–3032. [33] S. Arashloo and J. Kittler, “Energy normalization for pose-invariant face recognition based on MRF model image matching”, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 33, no. 6, (2011), pp. 1274– 1280. [34] R. Weng, J. Lu and Y.-P. Tan, “Robust Point Set Matching for Partial Face Recognition”, IEEE transactions on image processing, vol. 25, no. 3, (2016). [35] B.F. Klare, B. Klein, E. Taborsky, A. Blanton, J. Cheney, K. Allen, P. Grother, A. Mah, M. Burge and A. K. Jain, “Pushing the Frontiers of Unconstrained Face Detection and Recognition: IARPA”, IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , (2015), pp.7-12. [36] M. Tistarelli, Y. Sun and N. Poh, “On the Use of Discriminative Cohort Score Normalization for Unconstrained Face Recognition”, IEEE Transactions on information forensics and security, vol. 9, no. 12, (2014). [37] Y. Gao and H. Jong Le, “Pose Unconstrained Face Recognition based on SIFT and Alignment Error”, IEEE International Conference on Audio, Language and Image Processing (ICALIP) , (2014), pp.277281. [38] L. Best-Rowden, S. Bisht, J. C. Klontz and A.K. Jain, “Unconstrained Face Recognition: Establishing Baseline Human Performance via Crowd sourcing”, IEEE International Joint Conference on Biometrics (IJCB), (2014), pp. 1 – 8. [39] J. Ylioinas, A. Hadid, J. Kannala and M. Pietikainen, “An In-depth Examination of Local Binary Descriptors in Unconstrained Face Recognition”, 22nd International Conference on Pattern Recognition (ICPR), (2014), pp. 4471-4476. [40] V. Jain and E.Learned-Miller, “FDDB: A Benchmark for Face Detection in Unconstrained Settings”, University of Massachusetts-Amhert, (2010). [41] S. Sengupta, J.-C. Chen, C. Castillo, V. M. Patel, R. Chellappa and D. W. Jacobs, “Frontal to Profile Face Verification in the Wild”, IEEE Winter Conference on Applications of Computer Vision (WACV) , (2016), pp. 1-9. [42] J.-C. Chen, V. M. Patel and R. Chellappa, “Unconstrained Face Verification using Deep CNN Features”, IEEE Winter Conference on Applications of Computer Vision (WACV), (2016), pp.-1-9. [43] S. Liao, Z. Lei, D. Yi and S. Z. Li, “A Benchmark Study of Large-scale Unconstrained Face Recognition”, IEEE International Joint Conference on Biometrics (IJCB), (2014), pp. 1-8. [44] D. Retno Anggraini, “Face Recognition Using Principal Component Analysis and Self Organizing Maps”, Third ICT International Student Project Conference (ICT-ISPC-2014), (2014), pp.91-94. [45] S. Liao, A. K. Jain and S. Z. Li, “A Fast and Accurate Unconstrained Face Detection”, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 38, issue 2, (2016), pp.211-223. [46] C. Indhumathi, “Unconstrained Face Recognition From Blurred and Illumination with Pose Variant Face Image Using SVM”, International Journal of Innovative Research in Computer and Communication Engineering, vol.2, Special Issue 1,(2014), pp. 2564-2567. Copyright © 2016 SERSC 265 Bo ok O m nl ad in e eV by e th rsio is n fil O e is nly IL . LE G AL . International Journal of Hybrid Information Technology Vol. 9, No.11 (2016) 266 Copyright © 2016 SERSC