The preliminary study of the face recognition technique reflects the set of algorithms for the process of face recognition. The study evaluates that the recently the face recognition technique is acquiring much attention in the society for providing safety and enjoying the other multimedia operation through this. The research work analyses the different state of the face recognition algorithm to select the best algorithm for the utility purpose. In the addition the research associate emphasises on the 3-dimendional face recognition approach for determining the size, position, pose of the head. The 3-dimensional model also provides more accuracy in the face recognition technique. The study identifies that the biometric facial recognition method is the most promising approach to recognise the individual’s face. Furthermore, the study also speculates the current situational challenges of the face recognition technique. Finally, the application of facial recognition in several domains has been addressed in the present research.
In case of social interaction, human face plays a crucial role in conveying the people’s identity.Considering the view of Ani (2012), biometric face recognition technology has received a significant attention by using the human face for the security purpose. Comparing with other biometrics system like fingerprint or palm print, face recognition has distinct advantages for its non-contact process. The face images for the recognition purpose can be confined from a long distance without touching the person and the identification does not require any kind of interaction with the person. Furthermore, face recognition serves as a crime restriction principle as the recorded face image can help to identify the person later (Bate and Lecturer, 2012).Over the past decades several scientific research centres have concentrated on developing the face recognition methods within the framework of biometric security systems. Now a day, the several security departments in hotel industries, police department and the other government offices have equipped the facial recognition technology for making the safety system (Wang and Wu, 2010).The purpose of the research is to review the algorithm of the face recognition technology. Along with that the research associate also analyses the 3-dimensional and biometric facial recognition method in the present days. Furthermore, the challenges of using this technology and the deployment examples to mitigate the issues are also addressed in the study.
Face Recognition Algorithms:
The two important metrics such as geometry and photometry are two initial parameters for the development of facial recognition algorithm. For implementing these approaches wide range of algorithms has been developed by the scientist. The research has emphasised on the three significant torrents of work such as Principal Components Analysis (PCA), Linear Discriminant Analysis (LDA) and the Elastic Bunch Graph Matching (EBGM) (www.animetrics.com, 2016). According to the opinion of (Bhowmik, 2012), the PCA technique can convert the two dimensional image into one dimensional vector. This feature actually selects the features of the images. The prime advantages of the PCA technique are that it uses eigenface approach which helps to reduce the size of the database for test images recognition. LDA is a statistical approach based on a set of training images of known people. In the opinion of Dong and Chen (2014), this technique helps to find out the underlying vectors in the facial feature space and eliminates the variance within a number of sample images of the same person. This algorithm assists to recognise the actual person,despite having the image discrimination of the same individual. The EBGM depends on the concept of the real image of a particular person. The EGBM technique spaces for small blocks of numbers named as a Gabor filter over the smallerareas of the image (Hsieh and Tung, 2009). The process of the algorithm is to multiply and add the blocks with pixel value of the images to fabricate numbers at the various locations on the image. This is a new technique which greatly enhanced the performance of the facial recognition method under the variety of expression, angle and pose.
The other two representatives for the face recognition algorithm are the eigenface approach and neural network. The eignface approach represents the Karhonen-Loeve (KL) transform for the feature extraction (Martensen, 2015). This algorithm eliminates the facial feature dimension along with that maintain the reasonable discriminating power of the image. In this technique, each face is represented as a linear combination of the eigenfaces to make the picture more perfect. The algorithm follows the steps to reduce the discrimination.
It acquires database for the face image to determine the free spaces within the image.
After recognising a new image the algorithm calculates the weights.
It determines the free image spaces to reduce the obstruction (Sharma and Patterh, 2015).
Finally, it tries to match the corresponding image with the face from the database.
On the other hand, the neural network approach provides the sophisticated modelling scheme for estimating the pattern recognition phase. To reduce the image complexity, the neural network is combined with the pattern recognition, image phase rather than the feature extraction phase. According to the opinion of Omidiora et al. (2009),another technique is line edge maps (LEM) to perform the face recognition method. LEM uses physiologic features to solve the face related problem of human that mainly occurs with mouth, nose and the eye area. This technique converts the images into grey level pictures for measuring the similarities of human faces. Then it encoded with binary edge map by using the Sobel edge detection algorithm (Pato et al. 2010). This algorithm is advantageous for its low memory space utility. This algorithm keeps the face features in a simplified level for easy recognition. The above discussionsheds the light on the algorithm of each category of the face recognition technique to identify the best alternative.
In the facial recognition software, the newly-emerging trend is to use the 3D model. Considering the view of Hu (2008), the 3D model claims the provision for more accuracy. 3d facial recognition applies the distinctive feature by capturing the real-time 3D image of any person’s facial surface. The rigid tissue and the bones are much apparent along with the nose, chin, curves of the eye socket. However, Huang et al. (2010)argued that these areas are totally unique and does not change over time. 3D facial recognition can also be used in the darkness by using the depth and axis of measurement. Moreover, this recognition process has also the ability for recognising the subject at different view angles for recognising up to 90 degrees. The system of 3D software includes various steps. The first phase is detection. Here an image is acquired and can be accomplished digitally through scanning the existing photograph (2D) or using the video image for acquiring the live picture of the 3D subject. The next step is alignment which detects a face. The system also determines the position, size and pose of the head. Kukharev et al. (2010) stated that the subject has the ability to be recognised up to 90 degrees while 2D is required to be turned at least 35 degrees toward the camera.
The next stage is measurement which measures the curves of any face on the sub-milimeter scale. It also creates the template. According to Ming (2015), the representation is the system which translates the template to the unique code. Such coding provides each template the numbers for representing the features on the subject’s face. Finally, if the image is 3D, it matches with the database images. When the 3D image is usually taken, various points are identified.
The bio-metric based facial recognition techniquehas been emerged to be the most promising option to recognise individuals. This is used instead of authenticating the people and granting access to the virtual and physical domains depending on the smart cards, password, tokens, plastic cards and keys. The password and Pins are hard to remember and can also be forgotten, duplicated or purloined. On the other hand, Ryazanov et al.(2009) argued that an individual’s biological traits could not be misplaced, stolen, forged or forgotten. Considering the opinion of Wang and Shi (2009), the biometric-based facial recognition technologies include the physiologiocal characteristics like fingerprints, face, finger geometry, palm, hard veins, retina, voice, ear and iris. For example, Identix has developed the new product fo0r helping with precision. Furthermore, the development of FaceIT Argus utilises the skin biometrics, the skin texture’s uniqueness and yielding more accurate results. However, Xu and Chen (2009) mentioned that this technique of biometric face recognition is not the perfect system. Few challenging factors in the way of recognition are significant glare on the eyeglasses or wearing the sunglasses (Rhodes et al. 2009). The long hair also obscures the central part of the face during facial recognition. On the other hand, the poor lighting could cause the face to be under or over exposed. Moreover, the lack of resolution can be the challenging issue during the biometric facial recognition process. Good face recognition algorithms and the appropriate processing of the images can hugely compensate the noise variations in scale, illumination and orientation.
The problem in face recognition arises as the standard pattern classification or the machine learning problem. The facial ageing, facial marks, face recognition in video and the near-infrared face recognition are the crucial challenge of the face recognition technology. As the age grows, the face shape or texture also changes. Considering the opinion of Ryazanov et al. (2009), the present face recognition engines are not as robust that these facial changes can be incurred from the ageing process. On the other hand, the facial marks support the textual retrieval of the candidate face images. In case of the 3D pose, the foreground-background segmentation and the illumination have become the pertinent issues for performing the facial recognition. On the other hand, the human face is not any unique object. According to Ming (2015), there are various faces and each of which is able to assume the variety of deformations. The inter-personal variations in face can be occurred due to the identity, race or the genetics while the intra-personal variations can be raised due to the expressions, deformation, facial hair, ageing, facial paraphernalia and cosmetics. A recognition system is required to associate a name or identity for every face which comes across through matching it with large database of individuals (Xu and Chen, 2009). Hence, the system is required to be robust for the typical image-acquisition issues, like video-camera, image resolution and noise. Hence, the fundamental issue is there with the recognition problem and multi-dimensional detection. The final constraint of the facial recognition technique is the requirement to maintain the usability of such system on the contemporary computational devices (Ming, 2015). The processing involved with such system would not be so efficient in regards to storage space and the run-time.
The face recognition technique establishes the vital presence of an authorised person rather than only just checking the valid identification. According to Ryazanov et al. (2009), for eliminating the duplicates in the nationwide voter registration process, the face recognition technique is used. It compares the face images of the voters along with the persons who do not use any ID number for differentiating one from others (Xie and Lam, 2008).
In many offices for the access control application, the face recognition is used if the size of the group of people is small. In such cases, the face pictures are caught under natural conditions, like indoor illumination and frontal faces. The face recognition in such system is done with high accuracy with much co-operation from the user (Huang et al.2010). In this case, any other user who attempts to login without the proper authorisation is totally denied. On the other hand, today the security has been the prime concern at the airports, airline staffs and the passengers. In this regard, the face recognition technology can be used to overcome the crimes.
The primary assertion of the study reflects the process of face recognition technique in the security department. The study revealed that face recognition is the unique process of recognition technique as well as it has numerous challenges to deal with the present age. The research associate has been focused on the advantages of face recognition technique among the other biometric technique. The research work has given an introductory survey of the face recognition technology. The entire research work has covered the area of 3-dimensional recognition technique, biometric facial recognition process and the challenges of the technology. The study also evaluated the several states of the face recognition algorithms and the advantages of several algorithms. The research also provides the process to mitigate the weakness of the facial recognition system. Furthermore, it has been speculated that there is some cost effective and reliable technology of face recognition method. As a result, there are no financial barriers to the widespread deployment of the face recognition technology. Finally, the study has analysed the application of facial recognition techniques in different level like official purpose, voter card and the other security system
Ani, M.S. Al (2012) ‘Face recognition approach based on Wavelet – Curvelet technique’, Signal & Image Processing : An International Journal, 3(2), pp. 21–31.
Bate, S. and Lecturer (2012) Face recognition and its disorders. Basingstoke: Palgrave Macmillan.
Bhowmik, M.K. (2012) ‘Eye region-based fusion technique of thermal and dark visual images for human face recognition’, Optical Engineering, 51(7), pp. 205–210.
Dong, X. and Chen, H. (2014) ‘An improved face recognition technique based on modular multi-directional Two-dimensional principle component analysis approach’, Journal of Software, 9(12), pp. 122–128.
Facial recognition applications – Animetrics, Inc (2012) Available at: http://animetrics.com/technology/frapplications-html/ (Accessed: 12 April 2016).
Hsieh, P.-C. and Tung, P.-C. (2009) ‘A novel hybrid approach based on sub-pattern technique and whitened PCA for face recognition’, Pattern Recognition, 42(5), pp. 978–984.
Hu, H. (2008) ‘ICA-based neighborhood preserving analysis for face recognition’, Computer Vision and Image Understanding, 112(3), pp. 286–295.
Huang, H., He, H., Fan, X. and Zhang, J. (2010) ‘Super-resolution of human face image using canonical correlation analysis’, Pattern Recognition, 43(7), pp. 2532–2543
Kukharev, G., Kamenskaya, E., Electrotechnical, S.-P. and gkukharev@ (2010) ‘Reference tools’, Pattern Recognition and Image Analysis, 20(2), pp. 210–219.
Martensen, K.M. (2015) Radiographic image analysis. United States: Elsevier – Health Sciences Division.
Ming, Y. (2015) ‘Robust regional bounding spherical descriptor for 3D face recognition and emotion analysis’, Image and Vision Computing, 35, pp. 14–22.
Omidiora, E., Fakolujo, O., Ayeni, R. and Ajila, T. (2009) ‘A survey of face recognition technique’, Journal of Applied Science, Engineering and Technology, 7(1), pp. 226–230.
Pato, J.N., Millett, L.I., Committee, W.B., Science, C., Board, T., Division, Sciences, P., Council, N.R., Computer Science and Telecommunications Board and Division on Engineering and Physical Sciences (2010) Biometric recognition: Challenges and opportunities. 2nd edn. Washington, D.C.: National Academies Press.
Rhodes, G., Locke, V., Ewing, L. and Evangelista, E. (2009) ‘Race coding and the other-race effect in face recognition’, Perception, 38(2), pp. 232–241.
Ryazanov, V.V., Vinogradov, A.P., Institute, M.S., Engineering, R. and Electronics (2009) ‘Reference tools’, Pattern Recognition and Image Analysis, 19(1), pp. 78–83
Sharma, R. and Patterh, M.S. (2015) ‘A face recognition system using PCA and AI technique’, International Journal of Computer Applications, 126(6), pp. 30–37.
Wang, T. and Shi, P. (2009) ‘Kernel Grassmanniandistances and discriminant analysis for face recognition from image sets’, Pattern Recognition Letters, 30(13), pp. 1161–1165.
Wang, Y. and Wu, Y. (2010) ‘Face recognition using Intrinsicfaces’, Pattern Recognition, 43(10), pp. 3580–3590.
Xie, X. and Lam, K.-M. (2008) ‘Face recognition using elastic local reconstruction based on a single face image’, Pattern Recognition, 41(1), pp. 406–417.
Xu, D. and Chen, X. (2009) ‘Method of modular 2DLPP and its application in face recognition’, Journal of Computer Applications, 29(8), pp. 2056–2059.