IJRIT International Journal of Research in Information Technology, Volume 2, Issue 4, April 2014, Pg: 267- 273

International Journal of Research in Information Technology (IJRIT)

www.ijrit.com

ISSN 2001-5569

Face Recognition in Surgically Altered Faces Using Optimization and Dimensionality Reduction Methods R.Ramya Priya 1, DR L.M Nithya2 1 2

( Department of Information Technology, SNS College of Technology, India) ( Department of Information Technology, SNS College of Technology, India)

ABSTRACT-Face recognition presents a demanding and challenging problem in the field of image analysis and computer vision. To recognize faces after plastic surgery is still arduous and difficult task. Plastic surgery procedures will be used to intensify the facial appearance of an individual to get a younger or good look. Apart from cosmetic reasons, plastic surgery procedures are beneficial for patients suffering from several kinds of disorganization caused due to excessive structural growth of facial features or skin tissues. There is a possibility of attacking others isolation by offender. To change facial geometry and texture of face increases the interclass variability and therefore, matching post-surgery images with pre-surgery images becomes a difficult task for automatic face recognition algorithms. Difference in pose, expression, illumination, aging and disguise are considered as major challenges in face recognition and several methods have been proposed to address these challenges, multi-objective evolutionary algorithms are used to match face images before and after plastic surgery. In proposed system, PSO (particle swarm optimization) can be used to bring out better accuracy in altered faces by using lbest and gbest methods and also the work can be extended to find misclassifications discovered in face by means of assuming cost for each misclassification. To find misclassification, dimensionality reduction or subspace methods are used, can be achieved either by eliminating data closely related with other data in the class or combining data to make a smaller or lesser set of features. Eventually false positive rate can be calculated to find misclassified face. INDEX TERMS-Face Recognition, Plastic surgery ,Reduction methods

I INTRODUCTION Facial recognition is the ability to recognize people by their facial characteristics. The system can conduct facial database searches and perform one-to-one, one-to-many verifications. Face recognition software or system that can be used to automatically identify or verify individuals from video frame or digital images. Facial recognition software is also known as a facial recognition system or face recognition software. All biometric recognition systems are susceptible to accidental errors of two types which both must be minimized. False accept (FA) errors where a random imposter is accepted as a legitimate users and False Reject (FR) errors where a legitimate users is denied access. Face appearance representation schemes can be divided in to local and global methods, depending on whether the face is represented as a whole or entire, or as a series of small regions. Normally the face recognition procedure or steps can be characterised as follows, they are Face Detection, Feature Extraction, and Face Recognition for recognizing correct face or legitimate user. The Plastic surgery procedures endure people and become more popular worldwide because of more important factors such as available of advance technology, the speed with which the procedures are performed and mainly of an affordable cost.The plastic surgery procedures completely changes the texture and shape of facial features The plastic surgery are of two categories, local Surgery: The local surgery mainly used for correcting abnormality or improving texture of a skin in a face. Examples of local plastic surgery includes chin, forehead, eyelids, correcting jaw, teeth structure. However local surgery leads to changes in facial feature but the overall appearance of face will be look similar to the original face. Global Surgery: It completely changes the facial structure which is known as full face lift. This type of surgery will be aimed for reconstructing the features in a face rather than simply to improve aesthetics. As of today different types of plastic surgeries are available like rhinoplasty which is used to change the shape of nose, blepharoplasty, cosmetic surgery and nonsurgical procedures like laser resurfacing. The statistics shows that besides, all age groups are interested in plastic surgery procedures, but sometimes plastic surgery procedures can be misused by individuals, who hide the identity of genuine user and unintentionally reject the genuine users. A prevalence has happened in china that a group of women has been terminated, because all of them have undergone facial plastic surgery and they all R.Ramya Priya, IJRIT

267

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 4, April 2014, Pg: 267- 273

unidentified. The custom officers failed to identify them in their existing passport picture. Moreover, it is expecting that more people will undergo plastic surgery in the near future. The goal in image analysis is to extract needed information for solving application-based problems.

Fig 1:Relation among plastic surgery, aging, and disguise variations

II RELATED WORK R. Singh, M. Vatsa, H.S. Bhatt describes the feature matching problem becomes major issues in the several or many face recognition system, novel approach customized to deal with the challenges of matching faces across variations caused by plastic surgeries. Facial plastic surgeries are typically performed either locally to correct defects, anomalies or to improve general skin texture. For e.g. to correct defects such as cleft lip and palate to improve nose structure, chin, etc., or globally to reconstruct the complete facial structure. Holistic approaches to face matching characterize entire face as one entity. Therefore, a part wise approach that is based on the intuition that appearance of one or more facial features may not change much across plastic surgery procedures. In such a part-wise framework, the approach exploits recent successes of sparse representations for face matching[1]. M. De Marsico, M. Nappi, D. Riccio describes analysed several types of global plastic recognition The nonlinear variations produced by surgical procedures are difficult to address with current face recognition algorithms. Developed an approach to integrate information derived from local regions to match pre and post-surgery face images .Observed that with respect to plastic surgery, more than one facial region may be affected due to a surgery procedure. For example, blepharoplasty which is primarily performed to amend forehead also affects eye-brows. Further, observed that large variations appears in the appearance, texture and shape of different facial regions, it leads to difficult for face recognition. Algorithms are used to match a post-surgery face image with pre-surgery face images. Few face recognition approaches may not provide mechanisms to added the concurrent variations introduced in multiple features. They recognize faces using a combination of holistic approaches together with discrete levels of information. Face recognition under pose, illumination and expression is meticulously studied by researchers and many techniques have been proposed to cater to these variations. They further geometrically normalized sketches and photos to match them through Eigen analysis[2]. B. Heisele, P. Ho, J. Wu and T. Poggio describes a semi-automatic alignment step in combination with support vector machine (SVM) classification was examined. Due to self-occlusion, automatic alignment procedures will eventually fail to compute the correct correspondences for large pose deviations between input and reference faces. Combining view-specific classifiers has also been applied to face detection. A probabilistic approach using part-based matching has been used for expression invariant and occlusion tolerant recognition of frontal faces. There are two global approaches and a component-based approach to face recognition and evaluate their robustness against pose changes. The first global method consists of a straightforward face detector which extracts the face from an input image and propagates it to a set of SVM classifiers that perform the face recognition. By using a face detector achieves translation and scale invariance [3]. Russell C.Eberhart (2006) emphasis that the dissimilar Evolutionary computation paradigms to operators each and every methods to focusing on how each affects search behaviour in the problem space. PSO consists of processing elements called particles in which each particle represent a candidate solution. PSO shares many similarities with evolutionary computation techniques such as GA's. The focus is on how each operator affects the paradigm’s behaviour in the problem space. There are many ways to implement a genetic algorithm (GA). The genetic algorithm operators are selection, crossover and mutation will examined and compare them with PSO operators. In PSO, the global best particle R.Ramya Priya, IJRIT 268

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 4, April 2014, Pg: 267- 273

found among the swarm is the only information shared among particles. The system is initialized with a population of random solutions and searches for optima by updating generations. The search process utilizes a combination of deterministic and probabilistic rules that depend on information sharing among their population members to enhance their search processes. However, unlike GA's, PSO has no evolution operators. Each particle in the search space evolves its candidate solution over time, making use of its individual memory and knowledge gained by the swarm as a whole. Compared with GAs, the information sharing mechanism in PSO is considerably different. The chromosomes share information with each other so the whole population moves like one group towards an optimal area in GA. The global best particle found among the swarm is the only information shared among particles in PSO. Meanwhile it is a one-way information sharing mechanism. GA, crossover occurs between (usually) randomly selected parents. In PSO, a particle does not explicitly exchange material with other particles. A vital difference is that a given particle is exerted only by its own previous best position and that of the best position in the neighbourhood or in the global population. In the local version of PSO, a particle is exerted only by one of its topological neighbours. Even better performance can be achieved by reducing the value of w during a run [4]. Anil K. Jain, Arun Ross and Salil Prabhakar (2004) says that Face recognition is a intruding method, and facial images are probably the most common biometric one used by humans to make a personal recognition. The applications of facial recognition in real time range from a static, a dynamic, uncontrolled face identification in a cluttered or occluded background (e.g., airport). The most popular approaches to face recognition are based on either: 1) the location and shape of facial attributes such as the eyes, nose, lips and chin, and their spatial relationships, or 2) the overall (global) analysis of the face image that represents a face as a weighted combination of a number of canonical faces. These above mentioned systems also have difficulty in recognizing a face from images captured from two drastically different views and under different illumination or lighting conditions. It is questionable whether the face itself, without any contextual information (concise explanation), is a sufficient basis for recognizing a person from a large number of identities with an extremely high level of confidence. In order to work well of a facial recognition system, it should automatically: to detect whether a face is present in the acquired or correct image, find the face if there is one and recognize the face from a general viewpoint (i.e., from any pose)[5]. Shuicheng Yan,Dong Xu, Hong-Jiang Zhang, Qiang Yang, and Stephen Lin emphasis that to provide insights into the relationship among the state-of-the-art dimensionality reduction algorithms, as well as to facilitate the design of new algorithms. A general framework known as graph embedding, has been proposed to provide a unified perspective for the understanding and comparison of many popular dimensionality reduction(subspace) algorithms. Moreover, the graph embedding framework can be used as a general platform to develop new algorithms for dimensionality reduction. A proposed a novel dimensionality reduction algorithm called Marginal Fisher Analysis by designing two graphs that characterize the intraclass compactness and the interclass separability, respectively, and by optimizing their corresponding criteria based on the graph embedding framework. This new algorithm is shown to effectively overcome the data distribution assumption of the traditional LDA algorithm. Thus, MFA is a more general algorithm for discriminant analysis[6].

III EXISTING SYSTEM The conventional or automatic face recognition algorithms will find difficult to recognize faces. The facial information will be used by face recognition algorithms either in a non-atomistic or extract features and process them into parts. In traditional method, GA used to recognize faces after post-surgery. Therefore optimisation algorithms are used for recognizing face after plastic surgery. GA consists of chromosomes are binary string of finite length. GA consists of three parameters such as selection, crossover and mutation. The selection refers to mechanisms for selecting individuals for reproduction according to their fitness. The crossover relates the method of merging the genetic information of two individuals. The mutation is realized as a random deformation of binary strings with a certain probability. Besides, feature selection can be performed i.e., GA acts as feature selector and fitness value can be calculated. By assigning different weights in face thereby fitness value can be calculated. The SVM acts as a classifier for classification. Pseudo code for a GA procedure Begin; Produce random population of P solutions (chromosomes); For each individual i2P, calculate fitness (i); R.Ramya Priya, IJRIT

269

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 4, April 2014, Pg: 267- 273

For iZ1 to number of generations; Randomly choose an operation (crossover or mutation); If crossover; Choose two parents at random ia and ib; Generate on offspring ic Zcrossover (ia and ib); Else If mutation; Choose one chromosome i at random; Generate an offspring ic Zmutate (i); End if; Calculate the fitnessfunction of the offspring ic; If ic is superior than the worst chromosome then substitute the worst chromosome by ic; Next i; Check if termination Z true; End;

IV PROPOSED SYSTEM PSO is a stochastic or random optimization technique based on the movement and intelligence of swarms or particles. PSO applies the concept of social interaction for solving a problem. It was developed in the year 1995 by James Kennedy (social-psychologist) and Russell Eberhart (electrical engineer). It uses a number of agents (particles) that constitute a swarm moving around in the high dimensional search space (constituent structure of all space) looking for the best solution. Each particle is treated as a singlepoint in a N-dimensional space which adjusts its “flying” according to its own flying experience as well as the flying experience of other particles in N-dimensional space.In order to improves face recognition, PSO can be used .The pso is an optimisation algorithm, which generates good results while recognizing face, with minimum selection of features. The information sharing is different in PSO, while comparing to other optimisation algorithms. The sharing in PSO is one way information sharing. Each particle keeps track of its coordinates in the solution space which are associated with the best solution (fitness) that has achieved so far by that particle. This value is called pbest. The other best value that is tracked by the particle is the best value obtained so far by any particle in the neighborhood of that particle. This value is called gbest. The basic underlying concept of PSO lies in accelerating each particle toward its pbest and the gbest locations. Pseudo code for a PSO procedure Begin; Produce random population of N solutions (particles); For each individual i2N: calculate fitness (i); Initialize the value of the weight factor, u; For each particle; Set lBest as the best position of particle i; If fitness (i) is better than pBest; pBest(i) Zfitness (i); End; Set globalBest as the best fitness of all particles; For each particle; Calculate particle velocity; Update particle position; End; Update the value of the weight factor, u; Check if termination Z true; End; A. Dimensionality Reduction Methods In addition to that, misclassification are need to be find or identified. That is different misclassification leads to different costs in face. In addition to that, dimensionality subspace analysis mechanisms can be used to achieve a R.Ramya Priya, IJRIT

270

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 4, April 2014, Pg: 267- 273

minimum overall recognition loss by performing recognition in the low dimensional subspaces. The main aim need to minimize total cost. Misclassification Total Cost= Misclassification + Test Cost B. Linear Discriminant Analysis and Marginal Fisher Analysis Along with that, dimension reduction or subspace derivatives such as LDA, MFA should be used based on costs. The main purpose of cost sensitive to minimize the total costs. In general aiming to prevent disasters caused by mistakes with a large cost. The misclassification cost can be categorized into two like class dependent cost and example dependent cost. The cost sensitive subspace analysis methods are called as feature reduction mechanisms. Linear discriminant analysis (LDA) is a classical statistical approach for supervised dimensionality reduction and classification. LDA computes an optimal transformation (projection) by minimizing the within-class distance and maximizing the between-class distance. The LDA method tries to find the subspace that best discriminates different face classes. It is achieved by maximizing the between class matrix sb ,while minimizing the within class scatter matrix sw in the projective subspace. The MFA is a linear method, in which interclass margin can better characterize the separability of different classes than LDA. The high dimensional data with hundred to tens of thousands of features and also with many irrelevant and redundant features. Confusion matrix can be calculated. The subspace analysis approach to explicitly extract costsensitive features for cost-sensitive face recognition. C.

Feature reduction using Cost-Sensitive Approach

In this module, extracted features dimension can be reduced by using dimensionality reduction methods such as cost-sensitive linear discriminant analysis (CSLDA) and cost-sensitive marginal fisher analysis (CSMFA) methods: D. Cost-sensitive linear discriminant analysis (CSLDA) Let W= [w1,w2,…wk] be the low-dimensional subspace of CSLDA to be sought, and W maps xi into a low dimensional feature yi, where yi=WTxi,yi . To characterize the cost information of each class, design the following two rules to compute the cost-sensitive within-class scatter matrix and the cost-sensitive scatter matrix of CSLDA: •

The higher the cost of samples in the qth class, themore importance they are, and larger weights are assigned to these samples to estimate the within-class variance of CSLDA. The larger the cost of misclassifying the samples in the qth class as the pth class, q , the higher weight is used estimate the between-class variance of CSLDA.



To measure the cost of each class, define an importance function f(q) to compute the importance of the samples in the qth class, where 1 . Generally, there are many potential functions to define f(q) which is a monotone function of the cost C(q,p).

A simple linear function is as follows: f(q)= To measure different costs of misclassifying the samples from the qth class into the pth class,define another function g(q,p) to compute the cost, g(p,q)=C(p,q) Then,define the cost-sensitive betweenfollows:

class scatter Sb and cost-sensitive within-class scatter

Sw of CSLDA as

Sb= Sw= R.Ramya Priya, IJRIT

271

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 4, April 2014, Pg: 267- 273

Where

and

are the average samples in the pth and qth classes, respectively, l(xi) is the label of xi , is the cost of

misclassifying face samples of the pth class into the qth class CSLDA can be formulated as follows: = E. Cost-sensitive marginal fisher analysis CSMFA CSMFA first computes two locality graphs G1 and G2 to describe the geometrical information of the training samples = and = where

denotes the k-nearest neighbors of x ,l(xi) and l(xj) denote the class labels of

xi and xj, respectively, and k1 and k2are two empirically prespecified parameters to define the neighborhood size. Here G1 and G2 are two graphs to characterize the interclass separability and intraclass compactness of the training samples, respectively. The CSLDA and CSMFA to show how to exploit cost information of training samples to extract costsensitive features for face recognition. The two rules to compute the cost sensitive within-class scatter matrix and the cost-sensitive scatter matrix of CSLDA. The Toy example can be used for better illustrating misclassifications present in face.Cost-sensitive classification using Neural network used as a classifier. The multilayered feedforward network consists of many interconnected nodes – neurons. Neurons are organized into one output layer and one or more hidden layers. Usually, the sigmoid activation fuction f (x)=1/(1+e−Qx) is used to calculate the output of the neuron. Backpropagation is a specific technique for implementing gradient descent in weight space for a multi layered feed forward network The basic idea of this technique is to efficiently compute partial derivatives of an approximating function realized by the network. They are computed with respect to all the elements of the adjustable weight vector w for a given value of input vector x. The learning procedure basically feeds the input vector to the network, calculates the output vector o of the network and compares it to the desired output vector y. Based on the difference, the backpropagation procedure performs a gradient descent in the weight space. Normally, the backpropagation algorithm minimizes the squared error of the network where oi is the actual output of the i-th output neuron and yi is the desired output. The most straightforward way to reduce misclassification costs is to leave the learning procedure intact and modify only the probability estimates of the network during the classification of unseen (testing) examples. The probability P(i) that an example belongs to the class i is replaced with the altered probability, which takes in account the expected costs of misclassification

It favours the classes with higher expected misclassification costs. Note that only the probability estimate is changed. The actual output of the network is left intact, so the back propagation learning works as in the original case.

V CONCLUSION Sometimes automatic face recognition algorithms are failed to recognize faces. In proposed work the algorithms such as feature selection and subspace analysis methods (cost sensitive) are used for recognizing faces in a R.Ramya Priya, IJRIT

272

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 4, April 2014, Pg: 267- 273

better way. The optimization and LDA, MFA algorithms are found to produce better recognition results, while comparing to traditional methods. The cost can be used for LDA,MFA. Feature selection algorithm switching between the two basic extractors that is SIFT and EUCLBP/LBP and helps in encoding discriminatory information for each face granule. Recognize pre-surgery face for the corresponding post-surgery face. The face feature can be extracted. The feature selection and optimization task has to be done by Multi-objective PSO(particle swarm optimisation) algorithm. The PSO selects best features and calculates weight value(i.e. fitness value is calculated).The better accuracy can be achieved and find number of misclassifications in faces. The CSLDA and CSMFA methods to achieve a minimum overall recognition loss by performing recognition in the low learned dimensional spaces.

REFERENCES [1] R. Singh, M. Vatsa, H.S. Bhatt, S. Bharadwaj, A. Noore and S.S. Nooreyezdan , “Plastic surgery: A new dimension to face recognition”, IEEE Transactions on Information Forensics and Security, vol. 5, no. 3, pp. 441–448,(2010). [2] M. De Marsico, M. Nappi, D. Riccio and H. Wechsler, “Robust face recognition after plastic surgery using local region analysis”, in Proceedings of International Conference on Image Analysis and Recognition, vol. 6754, pp. 191–200,(2011), [3] B. Heisele, P. Ho, J. Wu and T. Poggio, “Face recognition: component based versus global approaches”, Computer Vision and Image Understanding, vol. 91, pp. 6–21,(2003). [4] R. C. Eberhart and Y. Shi,“Comparison Genetic Algorithms and Particle Swarm Optimization” Proc. 7th international Conference on Evolutionary Programming, pp. 611-616,(2006). [5] Anil K. Jain, Arun Ross, and Salil Prabhakar, “An Introduction to Biometric Recognition” IEEE Transactions on circuits and systems for video technology, vol. 14, no. 1,(2004). [6] Shuicheng Yan, Dong Xu, Benyu Zhang, Hong-Jiang Zhang, QiangYang and Stephen Lin , “Graph Embedding and Extensions: A GeneralFramework for Dimensionality Reduction” IEEE Transactions on pattern analysis and machine intelligence, vol. 29, no. 1,(2007). [7] H.S. Bhatt, S. Bharadwaj, R. Singh and M. Vatsa , “On matching sketches with digital face images”, in Proceedings of International Conference on Biometrics: Theory Applications and Systems, pp. 1–7,(2010), [8] D.-Q. Dai and H. Yan , “Wavelets and Face Recognition,” in Face Recognition, K. Delac and M. Grgic, Eds. ITech, Vienna, Austria, pp.558, (2007) [9] Rechenberg, I. J. M. Zurada, R. J. Marks II and C. Robinson, Eds (2000),” Evolution strategy, in Computational Intelligence: Imitating Life”, IEEE Press, Piscataway, NJ,2007.

R.Ramya Priya, IJRIT

273

Face Recognition in Surgically Altered Faces Using Optimization and ...

translation and scale invariance [3]. Russell C.Eberhart (2006) emphasis that the ... Each particle in the search space evolves its candidate solution over time, making use of its individual memory and knowledge gained by the swarm as a ... exchange material with other particles. A vital difference is that a given particle is ...

127KB Sizes 0 Downloads 293 Views

Recommend Documents

Recognizing Surgically Altered Face Images
Training is performed with images of 360 subjects from the plastic surgery face database. The perfor- mance is evaluated on post-surgery images from the.

Survey on Face Recognition Using Laplacian faces
1Student, Pune University, Computer Department, K J College Of Engineering and Management Research. Pune .... Orlando, Florida, 2002, pp.3644-3647.

Survey on Face Recognition Using Laplacian faces - International ...
Abstract. The face recognition is quite interesting subject if we see in terms of security. A system can recognize and catch criminals and terrorists in a crowd. The proponents of large-scale face recognition feel that it is a necessary evil to make

Rapid Face Recognition Using Hashing
cal analysis on the recognition rate of the proposed hashing approach. Experiments ... of the images as well as the large number of training data. Typically, face ...

Rapid Face Recognition Using Hashing
cal analysis on the recognition rate of the proposed hashing approach. Experiments ... of the images as well as the large number of training data. Typically, face ...

Face Recognition Using Eigenface Approach
which the person is classified by comparing its position in eigenface space with the position of known individuals [1]. The advantage of this approach over other face recognition systems is in its simplicity, speed and .... The experiment was conduct

Multithread Face Recognition in Cloud
Oct 12, 2016 - biometrics systems have proven to be an essential security tools, ... showcase cloud computing-enabled face recognition, which utilizes ...

Multithread Face Recognition in Cloud
Oct 12, 2016 - this responsibility to third-party vendors who can maintain cloud ... showcase cloud computing-enabled face recognition, which utilizes ...

Face Recognition in Videos
5.6 Example of cluster containing misdetection . .... system which are mapped to different feature space that consists of discriminatory infor- mation. Principal ...

Face Recognition using Local Quantized Patterns
by OSEO, French State agency for innovation and by the ANR, grant reference ANR-08- ... cessing, 2009:33, 2009. [18] H. Seo and P. Milanfar. Face verification ...

Face Recognition Using Uncorrelated, Weighted Linear ...
and within-class scatter matrices, respectively; m is the mean of all samples and mi is the mean of class .... For illustration, some available images for one subject ...

Face Recognition Using Composite Features Based on ...
Digital Object Identifier 10.1109/ACCESS.2017.DOI. Face Recognition Using Composite. Features Based on Discriminant. Analysis. SANG-IL CHOI1 ... ing expressions, and an uncontrolled environment involving camera pose or varying illumination, the recog

Face Recognition Using Sparse Approximated Nearest ...
The authors are with the School of Computer Science & Software. Engineering ..... adapting the Accelerated Proximal Gradient (APG) method to optimize (4) ...

Face Tracking and Recognition with Visual Constraints in Real-World ...
... constrain term can be found at http://seqam.rutgers.edu/projects/motion/face/face.html. ..... [14] Y. Li, H. Ai, T. Yamashita, S. Lao, and M. Kawade. Tracking in.

SURF-Face: Face Recognition Under Viewpoint ...
A grid-based and dense extraction of local features in combination with a block-based matching ... Usually, a main drawback of an interest point based fea- .... navente at the Computer Vision Center of the University of Barcelona [13].

SURF-Face: Face Recognition Under Viewpoint ...
Human Language Technology and ..... In CVPR, Miami, FL, USA, June. 2009. ... In International Conference on Automatic Face and Gesture Recognition,. 2002.

Face Recognition Based on SVM ace Recognition ...
features are given to the SVM classifier for training and testing purpose. ... recognition has emerged as an active research area in computer vision with .... they map pattern vectors to a high-dimensional feature space where a 'best' separating.

A 23mW Face Recognition Accelerator in 40nm CMOS with Mostly ...
consistently good performance across different application areas including face ... Testing on a custom database consisting of 180. HD images, 99.4% of ...

Improving Face Recognition in Real Time for Different head Scales
Abstract. Face recognition is technique which is used to identify person by their face. Identifying the person is performed with different types of Biometric like iris scan, finger print, face recognition, Gait recognition; Signature etc. face recogn

Improving Face Recognition in Real Time for Different head Scales
There are many problems are occurred when it perform face recognition like illumination, light intensity, blurred face, noisy image, tilted face, different head pose ...

Face identity recognition in autism spectrum disorders
Department of Brain and Cognitive Science and McGovern Institute for Brain ..... orientation (upright > inverted) was marginally significant (p = .054). T (CA). 15.

Appearance-Based Automated Face Recognition ...
http://sites.google.com/site/jcseuk/. Appearance-Based Automated Face. Recognition System: Multi-Input Databases. M.A. Mohamed, M.E. Abou-Elsoud, and M.M. Eid. Abstract—There has been significant progress in improving the performance of computer-ba

Markovian Mixture Face Recognition with ... - Semantic Scholar
cided probabilistically according to the probability distri- bution coming from the ...... Ranking prior like- lihood distributions for bayesian shape localization frame-.