20-04-2013, 04:21 PM
FACE RECOGNITION AND AUTHENTICATION
FACE RECOGNITION.doc (Size: 872 KB / Downloads: 29)
ABSTRACT
Human race is considered to be the superior most creation of nature. It is the only species, which has the ability to think effectively. Every human being is unique in this world and has his own characteristic features. But due to some similarity between persons, there is every possibility of mistaking one person for the other. This may not considerably affect in general public life. But when it comes to certain issues it cannot be overlooked. This presentation namely” FACE RECOGNITION AND AUTHENTICATION” is an attempt to eliminate this problem.
In this paper, we want to give a brief overview of Face authentication system integrating both color and range data based on low cost 3-Dsensor, capable of real time acquisition of 3-D and color images. Novel approaches are proposed that exploit depth information to achieve robust face recognition and authentication under conditions of background clutter, occlusion, face pose alteration and harsh illumination. The well known embedded hidden markov model technique for face authentication is applied to depth maps and color images. To cope with pose and illumination variations, the enrichment of face databases with synthetically generated views is proposed. The performance of the proposed authentication scheme is tested thoroughly on two distinct face databases of significant size.
INTRODUCTION
In the last 25 years, face authentication has received growing interest, in response to the increased number of real world applications requiring detection and recognition of humans, as well as the availability of low cost hardware. Although human faces have generally the same structure, they are at the same time very different from each other due to gender, race, and individual variations. In addition to these variations, facial expressions can change their appearance. The majority of face recognition techniques employ two-dimensional (2D) grayscale or color images. This is mainly due to the high cost of available 3-D digitizers that makes their use prohibitive in real-world applications. Furthermore, these devices often do not operate in real time (e.g., time of flight laser scanners) or produce inaccurate depth information (e.g., stereo vision). The work presented in this paper is partly motivated by the recent development of novel low-cost 3-D sensors that are capable of real-time 3-D acquisition. Another motivation comes from the fact that the 3-D structure of the face may be exploited to discriminate among individuals or aid 2-D face authentication, since the face shape data is not sensitive to variations of illumination, face pigmentation, and cosmetics, and is less sensitive to use or nonuse of glasses and facial expressions compared to 2-D surface reflectance data represented by 2D images.
Why face recognition?
• Speed. face enrollment time is less than 1 sec and matching speed is up to 65,000 faces per second in 1:N identification mode Multiple faces' processing.
• Detects all faces in the current frame and allows processing all of them.
• Doesn’t require any specific hardware. Face image can be obtained from low cost camera or web cam. Image processing and recognition are performed on standard PC.
ACQUISITION OF 3-D AND COLOR IMAGES
A 3-D and color camera capable of synchronous real-time acquisition of 3-D images and associated color 2-D images is employed. The 3-D data acquisition system is based on an active triangulation principle, making use of an improved and extended version of the well-known coded light approach (CLA) for 3-D data acquisition. The basic principle lying behind this device is the projection of a color-encoded light pattern on the scene and measuring its deformation on the object surfaces. The 3D camera achieves real-time image acquisition for range images and near real-time (14 fps) for color plus depth images on a standard PC. It is based on low-cost devices, an off-the-shelf CCTV-color camera, and a standard slide projector. By rapidly alternating the color-coded light pattern with a white-light pattern, both color and depth images are acquired. For the experiments in this paper, the system was optimized for an access control application scenario, leading to an average depth accuracy of 1 mm for objects located about 1 m from the camera in an effective working space of 60* 50* 50 cm. Computed depth values are quantized into 16 bits.
FACE RECONITION USING COLOR AND DEPTH IMAGES
Face recognition is most concerned with roughly finding all the faces in large, complex images, which include many faces and much clutter, localization emphasizes spatial accuracy, usually achieved by accurate detection of facial features. Several face detection techniques have been proposed for grayscale images. These may be roughly categorized to those based on the detection of facial features, possibly exploiting their relative geometric arrangement, and those based on the classification of the brightness pattern inside an image window, obtained by exhaustively sweeping the whole image as face or non face. Techniques in the second category were recently shown to be more successful in detecting faces in cluttered backgrounds; however, the correct detection rates reported were below 90%. Further shortcomings of existing face detection algorithms include their sensitivity to partial occlusion of the face (e.g., glasses and hair), hard illumination and head pose, as well as their computational complexity. However, the parameters of the color distribution were shown to rely on the environmental illumination and the response characteristics of the acquisition device. Furthermore, irrelevant skin colored image regions will result in erroneous face candidates.
ENROLLMENT
One of the main problems in face recognition is that facial appearance is distorted by, for example, seasonal changes (aging, hairstyle, usage of cosmetics, etc.), rotation, harsh or heterogeneous illumination, and occlusions caused by glasses, scarves, etc. This problem may be partly alleviated by recording a rich training database containing representative variations. Such an approach is shown to lead to improved recognition/authentication rates. A small set of images (normally less than five per person) depicting different facial expressions with and without eyeglasses are originally recorded. These are subsequently used to create canonical face images (upright pose). For each canonical image pair (color and depth), a 3-D model is constructed and used to automatically generate artificial views of the face depicting pose and illumination variations. The various steps of this procedure are described.
Surface Interpolation
In face authentication if some pixels are undetermined in the depth image, the corresponding pixels will also be missing in the artificially generated color images depicting various pose and illumination combinations. In practice, more than 20% of the face surface pixels are missing, mainly over occluded points on one side of the nose and face or over the eyes, for people wearing glasses. Therefore, a semi automated surface interpolation procedure has been developed that exploits face symmetry. Missing pixels from the right part of the face may be copied from symmetrically located pixels on the left part of the face, or vice versa. This procedure requires the estimation of the vertical bilateral symmetry axis with high accuracy.
CONCLUSION
Earlier we observed that there are a variety of different biometric authentication technologies, and these have different relative advantages and disadvantages in terms of cost, safety, psychological effect, accuracy, and so on. Among these, face authentication stands out as particularly user friendly with excellent acceptability on the part of users. In contrast to fingerprinting and iris matching methods that employ a mechanical device, face authentication involves much less psychological impact, and of course identifying people from their faces is an entirely natural human tendency. This was why face authentication was adopted for the "e-Airport Project" trials now in progress and why next-generation passports have adopted a facial data-based method