Face Image Quality - Report Example

PROPOSED PROJECT ON FACE IMAGE QUALITY By Name Course Instructor Institution Location Date Project Overview and its Potential Applications This paper is on a face image quality project that is to be developed with the main objective of achieving normalized face images. The techniques to be incorporated into the project will see it enable the users realize enhanced abilities of face recognition. Development of algorithms will be vital so as to enable easy and effective eye location detection. The algorithms will also be vital in the process of feedback provision to the user so that optimal performance is achieved through user position adjustment relative to the position of the camera. Testing of the interfaces and algorithms will occur with the help of images of faces retrieved from a database in order to ascertain the effectiveness of the algorithms and interfaces used. Issues Addressed in the Project The most basic issue that the project is to attempt is to link certain digital images to already pre-existing realistic images from databases. This can be inferred from the general understanding of face recognition as a process that uses computer applications which detect and verify human beings from the video sources or digital images. The project will work through comparison of the video resources or the digital images which contain facial characteristics fed into it with already stored images in the database. The project also seeks to achieve the ability to effectively achieve normalized images of faces. The project thus seeks to make fundamental improvements to already existing projects of quality face images with regard to the ISO specifications. The improvements will see the project tailored to achieve standard and quality face images. The project will use sets of algorithms and interfaces that will enhance image quality experience. The experience is enhanced by action of automatic recognition of faces and feedback with regard to positioning of users. Importance of the Face Image quality Project As a form of biometric technologies, the face image detection software which depends on the quality of the images present for investigation is not identified as the single most efficient and reliable [1]. However, its contributions to the biometric techniques remain unquestionable. The major strength of this project is the ability to give high quality face images without the necessary cooperation of the individual user whose facial features are under investigation in order to work effectively. Facial images quality determination systems appropriately designed and installed have the ability to capture images of faces from a reasonable distance and from among groups of people [2]. This is not the case with such technologies applied in iris and finger scans. Properly designed systems installed in airports, multiplexes, and other public places can identify individuals among the crowd, without passers-by even being aware of the system. Other biometrics like fingerprints, iris scans, and speech recognition cannot perform this kind of mass identification. However, questions have been raised on the effectiveness of facial recognition software in cases of railway and airport security. Theoretical Foundations upon Which the Project is Based The face image quality is determined by the algorithms that have been utilized to realize quality image technologies that enable them to extract specific features on the surface of the images in order to identify the faces [4]. The face whose image is to be taken relative to the camera position, the size of the image being scanned, eye size and shape, the organization of the jaw and the cheekbones and the nose are the key features that the algorithms will base its effectiveness upon[3]&[4]. To normalize facial images may require compression of information retrieved from study of faces and save details that are only important to the process of processing of quality face images[5]. This way, any face under investigation will only be tested for standard quality if it meets the set against the selected saved facial details. Initial facial technologies used techniques of matching templates in which sought to identify outward features of faces. This technology would then give forms of compressed face representations. Effective recognition which occurs when face images are scrutinized can be done in two major ways; photometric and geometric[6]. In photometric technique, an image to be scanned is distilled into numbers which are then subjected to comparison in which variant values are eliminated and matches are retained, proving a positive recognition. The geometric technique utilizes the distinguishing features of images to carry out face recognition. Past Attempts to Work on Face Image Quality Projects and the Results Obtained The FBI has made attempts to use high quality face images to achieve face recognition for security purposes[7]. In a project has been dubbed Computer-Aided Facial Recognition Project, the significance of face image quality cannot be overlooked. The FBI considered the need for the project against the protections and privacy risks incurred before authorization of the development of the project. The project was specifically designed to aid studies at the Sheffield University as a move towards enhanced statistical analysis of geometry of facial features. Collection of biographical and images of faces of willing individuals was done, with each of these individuals adequately informed on the significance of this study. A major database was created. The final database for facial images was delivered to the FBI agencies, and its design entailed a 2D description of images of users’ faces, biographical information about the users, including their family, names, ethnicity, age and sex, a 3D recording of information regarding the facial features of the users and a specific identification number. The project was adequately designed and the facial image quality role was noted. The geometry of facial landmarks was enhanced, with such techniques as determination of the distance between the mouth and nose and determination of ear tips done as a means of distinguishing from one person to the other for security purposes. The FBI thus has been able to compare and judge images of suspected individuals with the database images with the help of the generated statistics from the project through incorporation of a system for semi-automated facial features extraction tool. Proposed Solutions to Face Image Quality Problems Failure of achieving high quality facial images has infiltrated to facial recognition leading to criticism for the folly committed in mis-alignment[3]&[8]. It is for this reason that a new approach to mis-alignment has been proposed by Shiguang Shan in collaboration with four other authors to correct the problem noted in the process. They propose an initial stage of investigation of the problem using empirical resources in which systematic evaluation of the sensitivity of Fisherface with regard to mis-alignment through disturbance of the co-ordinates of the eye on the Feretface database. This process confirms Fisherface system degeneration due to facial landmarks with imprecise locations. The next step entails analysis of the sources of the problem noted in mis-alignment, followed by sub-categorization of the causes of the problem into alignment returning, mis-alignment modeling and invariant features. The authors then give a prospective way of making measurements with a combination of the rate of recognition with the error of distribution of the alignment which subsequently lead to a better evaluation of the general effectiveness of predetermined approaches of face recognition against the considered problematic mis-alignment. They finally propose the E-Fisherface method of learning mis-alignment which is reinforcement the recognition tool which manages the variations of mis-alignment. Secondly, face images through FLDA have the limitation of having a single image against one person[9]. The failure of FLDA (Fisher linear discriminant analysis) in this case is caused by the difficulty of calculation of scatter matrices of a within-class. However, the one person problem can be overcome by application of the image decomposition techniques. The first step entails placement of an image and its approximations within the training set. The date generated enables effective application of the 2 diLOCKIE, M. (2002). Biometric technology. Chicago, Heinemann Library.mension FLDA. Testing of the effectiveness of the image decomposition algorithm occurs on 5 variant face databases. This proposed solution is more effective than the SVD approach with regard to the training time and rate of recognition in the stated 5 databases. Works in Related Fields Applicable to Face Image Quality Project There exist other projects done that work in close resemblance as Face Image Quality project. Biometrics as a study is concerned with the analysis of human traits[10]. It is an application of computer science which attempts to access control and identify given information about people in groups undergoing investigation. Just like face image quality, biometrics identifies landmarks in individuals which are specific and measurable which can help in the process of describing andassigning of labels to people[11]. Identifiers of biometrics are subcategorized as either behavioural or physiological. The physiological features of individuals have a resemblance to face image quality as they have to do with the body shape. Illustrations of the physiological make up of an individual can entail their retina, iris recognition, hand geometry, palm print and face recognition. The traditional modes of control of access in biometrics may show slight variations from the current forms of face image quality assessment, as traditional control access means were inclusive of token-based systems of identification such as passports and systems of identification that are based on knowledge recovery as passwords. Just as software for face image quality identifies people distinctively so does biometrics since the identifiers have unique and distinctive features to people. These identifiers are more effective and reliable in the process of identifying individuals. Technology Used in Face Image Quality and Probable Impacts of the Proposed Project Face image quality utilized the 3D recognition as its basic technology[3]. 3 dimensions have led to enhanced image investigation. Three dimension sensors are the basic recorders of information specific to the users’ facial landmarks. The information accessed is then stored for use in identification of the distinctive traits on the faces of the users’ images, such as distance between their eyes and nose. The use of three dimension images has been noted as not affected by changes in the systems. Utilization of 3D capabilities in face recognition has seen identification of images from varying angles being a possibility. Three dimension information is more effective as it improves the quality of face images. The technology utilized sophistication of sensor development. The sensors are designed to project light onto the image or face under investigation. The project with 3D ability allows usage of up to twelve sensors, with each sensor playing a dedicated role of capturing information about specific facial landmarks; the sensors are mounted on a single CMOS chip. The project also uses the skin texture analysis technology which exploits the skin’s visual information obtained from digital images. Skin analysis technology has been designed with the ability to convert unique spots, patterns and lines on the skin under investigation into mathematical space[12]. Test on the effectiveness of usage of skin texture analysis have shown a five percent increase on accuracy and quality of face images. Limitations of Solutions to Face Image Quality Problems The effectiveness of face image quality is limited to frontal face appearances and a minimum of twenty degrees off[13]&[9]. An extension of the degrees above twenty reduces the effectiveness and accuracy of the project. The effectiveness of the project further is affected by adverse lighting conditions, and such foreign materials that might cover the surface of the face of an individual such as long hair and glasses. The facial expressions have been identified as a limitation to the accuracy and effectiveness of face quality projects. Frowning and smiling are instances in which face recognition tools might fail to give effective and desired results. For this reason, any database and investigated images need to be neutral and expressionless. Researchers and investigators using face image quality software have different databases. This inconsistency of databases reveals a weakness on the software, although the underlying problem is the use of varying numbers of stored pieces of information and images from one database to another. Proposed Technical Improvements Mei L. and Patrick J. have proposed technical improvements to the face recognition realized through utility of face image quality software project[15]. The improvements are expected to realize enhanced resources of computation and high definition image processing. The computation abilities to be improved target low quality webcam images and visa application images[14]. Such improvements could realize legally binding images as those used in driving licenses with the help of ISO certified images to reduce duplication of similar documents for illegal purposes. References [1]Karam, P. A. (2012). Artificial intelligence. New York, Chelsea House. http://public.eblib.com/choice/publicfullrecord.aspx?p=836166. [2] Li, S. Z., & Jain, A. K. (2005). Handbook of face recognition. New York, Springer. http://www.books24x7.com/marc.asp?bookid=16239. [3] Daoudi, M., Srivastava, A., & Veltkamp, R. C. (2013). 3D face modeling, analysis & recognition. http://search.ebscohost.com/login.aspx?direct=true&scope=site&db=nlebk&db=nlabk&AN=587978. [4] Li, J.-B., Chu, S.-C., & Pan, J.-S. (2013).Kernel learning algorithms for face recognition. http://dx.doi.org/10.1007/978-1-4614-0161-2. [5] Li, S. Z., & Jain, A. K. (2011). Handbook of Face Recognition. London, Springer-Verlag London Limited. [6] Zhou, S. K., Chellappa, R., & Zhao, W. (2006). Unconstrained face recognition. Springer E-Books. New York, Springer. http://public.eblib.com/choice/publicfullrecord.aspx?p=324743. [7] Kisku, D. R., Gupta, P., & Sing, J. K. (2013). Advances in biometrics for secure human authentication & recognition. [8] Zhao, W., & Chellappa, R. (2006). Face processing advanced modeling & methods. Amsterdam, Elsevier / Academic Press. http://www.engineeringvillage.com/controller/servlet/OpenURL?genre=book&isbn=9780120884520. [9] Bate, S. (2013). Face recognition and its disorders. [10] Iciar (Conference), Campilho, A., & Kamel, M. (2012).Image analysis & recognition 9th International Conference, ICIAR 2012, Aveiro, Portugal, June 25-27, 2012. Proceedings. Part II Part II. Berlin, Springer. http://dx.doi.org/10.1007/978-3-642-31298-4. [11] Campisi, P. (2013). Security & privacy in biometrics. London, Springer. http://search.ebscohost.com/login.aspx?direct=true&scope=site&db=nlebk&db=nlabk&AN=604387. [12] Zhang, Z. (2008). Integrating facial expressions & skin texture in face recognition. Thesis (Ph.D.)--State University of New York at Buffalo, 2008. http://proquest.umi.com/pqdweb?did=1594491201&sid=1&Fmt=2&clientId=39334&RQT=309&VName=PQD. [13] Lu, J. (2004). Discriminant learning for face recognition. Thesis (Ph. D.)--University of Toronto, 2004. [14] Lockie, M. (2002). Biometric technology. Chicago, Heinemann Library. [15] Wang, P. S.-P. (2011). Pattern recognition, machine intelligence and biometrics. Beijing, Higher Education Press. http://public.eblib.com/choice/publicfullrecord.aspx?p=885143. Read More