WE ARE ALL FACE PERCEPTION EXPERTS

Face Imperception or a portrait of prosopagnosia

The act of recognizing a face is actually quite complex. Like many visual stimuli, faces must be accurately recognized in any orientation or lighting condition, and even while moving. But unlike other objects, faces are intimately involved in communication, and our brains must be able to extract a tremendous amount of subtle detail from just a glance. So while some of the issues involved in face recognition are the same as for recognizing any object, other issues are unique to faces. Confronted with this dilemma, is the brain's most efficient solution to have special mechanisms for face recognition, or to simply extend the abilities of existing object recognition mechanisms?

That question is at the heart of a deep controversy in face recognition research. Studies with monkeys suggest that unique face recognition mechanisms might exist, while brain imaging experiments, studies with babies, and studies of people who cannot recognize faces show evidence supporting both sides of the argument. Compelling questions persist within the scientific community: what exactly are the mechanisms for processing faces in the brain? What compromises must be made for the brain to recognize faces quickly and accurately? And what does that indicate about how the human brain functions in general?

Faces enter the human visual experience from the very beginning of life. Studies show that as soon as nine minutes after birth, babies prefer to look at pictures that most resemble human faces. Moreover, young infants have great propensity for mimicking the facial gestures of people around them. Although infants practice many motor skills, these studies show that even at a very early age, humans devote a great deal of attention and energy to the movements required for facial expression. By adulthood, our infantile preferences translate into an ability to recognize human faces better than other visual stimuli.

Faces are the most important parts in recognizing individuals and facial expressions are very essential for human interaction and communication. You can read pleasure or displeasure, delight or frustration through facial expressions. Face carries a very great deal of interesting and important information for the brain. A long-standing goal of the face processing research agenda has been to identify which cues are extracted from a face in order to categorize it ( according to its gender, expression, identity, race, etc. ) Humans are generally considered face processing experts because they efficiently extract the diagnostic cues allowing face categorization, identification, and generalization.

Studies in humans using functional magnetic resonance imaging (fMRI) and electrophysiological equipment report neural activity uniquely related to viewing faces. These studies suggest that the right hemisphere of the brain may be more specialized to process faces than other objects in the visual field.

In 1997, Nancy Kanwisher and colleagues at Massachusetts Institute of Technology used fMRI to record the brain activity of subjects who were shown a series of faces and common objects. Their research showed that the fusiform gyrus became significantly more active when the subjects were presented with faces than when they were looking at other objects.

On another perceptual expertise test people were asked to recognize the object with the first name that comes to mind.

“Identify this object with the first name that comes to mind.”

EXPERT HAS INCREASED ACTIVATION OF FUSIFORM GYRUS. Fusiform Face Area” (FFA) is apparently specialized for face processing.

Orientation is also another important factor in face recognition. We are better at recognizing upright faces than recognizing upright objects.

In 1999, Kanwisher conducted a different fMRI study that looked at brain activity while subjects viewed pictures of human faces and pictures of animals. Kanwisher reported that again the fusiform gyrus became significantly more active when subjects looked at human faces than when they looked at animals. Results from these studies suggest that the fusiform gyrus is a region of the brain specialized for processing faces.

The temporal lobe of the brain is partly responsible for our ability to recognize faces. Some neurons in the temporal lobe respond to particular features of faces. When the appearance of a face is changed, neurons in the temporal lobe generate less activity.

Here is an interesting experiment:

Do you recognize the famous people in the pictures below? It may be difficult for you to recognize these people when they are upside-down. To identify these people, move your mouse so the cursor is over each picture. This will flip the pictures right-side up.

http://faculty.washington.edu/chudler/java/faces.html

Orientation is also another important factor in face recognition. We are better at recognizing upright faces than recognizing upright objects.

Even though the face on the left is scrambled it still creates the illusion of Thatcher in below example.

Test your face perception skills with two interactive experiments; one shows the inversion effect, and the other demonstrates how edges, features and surface area affect the process of human face recognition.
http://www.brainconnection.com/topics/?main=fa/face-perception

What Does Prosopagnosia Tell Us?

The capacity to recognize individuals through their faces is sometimes lost as a result of damage to specific areas of the visual brain which is known as fusiform gyrus. This syndrome is known as a prosopagnosia. Although able to detect a face among objects (“face detection”) prosapognosic patients lose the ability to identify familiar faces, including famous persons, friends, and relatives, or even their own face.

The clinical and anatomical conditions of prosopagnosia have been of great interest to cognitive neuroscientists willing to clarify the neurofunctional mechanisms of normal face processing. Anatomical descriptions of prosopagnosia support the critical role of the right hemisphere in the occipitotemporal pathway of face processing. The double dissociations reported between the ability to perceive unfamiliar and familiar faces (Malone, Morris, Kay, & Levin, 1982), between the recognition of facial expression and facial identity (e.g., Tranel, Damasio, & Damasio, 1988; Bruyer et al., 1983), or between lip reading and face identification (Campbell, Landis, & Regard, 1986) have helped in isolating the different subfunctions in a cognitive architecture of face processing (Bruce & Young, 1986).

Despite the theoretical importance of studies of prosopagnosic patients, an important area of research remains largely unexplored. In descriptions of clinical cases, it is usually reported that prosopagnosic patients who rely on nonfacial cues to recognize people (gait, voice, clothes, etc.) still attend to and extract information from faces that is used to recognize people. Yet, these patients appear to have lost the ability to extract and/or build diagnostic representations of other people’s faces. What is then the nature of the facial information that brain-damaged prosopagnosic patients extract when processing faces? Answering this question would undoubtedly contribute to functionally characterizing the selective face impairment that is prosopagnosia, and from there on providing a better understanding of functional aspects of the normal, ‘‘expert,’’ face processing system. However, several limitations may hamper a better understanding of the functional aspects of face processing in prosopagnosia.

First, cases of prosopagnosia usually suffer from many low-level and high-level visual deficits besides their face impairments: loss of visual acuity, visual field defects (hemianopia) achromatopsia, or difficulties at general configural processing and object recognition. One class of prosopagnosic patients usually knows that they are looking at a face. They can commonly recognize the details such as nose, mouth, eyes and so on. On one of Zeki Semir’s studies a patient looks at himself and certainly see a face, with nose, eyes and mouth etc, but somewhat it is not familiar; it really could be anybody. Like all the other prosopagnosic patients he was unable to bind all the individual features together and come up with recognizable face. There is one patient who was not able recognize his physiotherapist, while being treated by his physiotherapist, suffered a stroke that targeted his fusiform gyrus. He knew that he was looking at her face and knew exactly who she was but would not recognize her face. It is not surprising to find that prosopagnosic patients commonly have to use other features, for example voice or the clothes to identify a particular individual. One patient said that he can not recognize his wife except by the sound of her voice.

On another study (Roberto Caldara1, Philippe Schyns1, Euge`ne Mayer2, Marie L. Smith1,

Fre´de´ric Gosselin3, and Bruno Rossion4) a 54-year-old woman who sustained a closed head injury in 1991. After several months of spontaneous recovery and neuropsychological reeducation, she was left with massive prosopagnosia, being unable to recognize famous and familiar people. Despite large occipital and occipitotemporal lesions her low-level vision is almost perfect, her visual acuity being 8/10 in both eyes (August 2003), with a full visual field, apart from a small left paracentral scotoma. She reads normally (although slowly) and, crucially, does not present any problem at object perception and recognition, even for subordinatelevel discriminations (Rossion et al., 2003). Her deficit truly appears to be restricted to the category of faces. With faces, she is able to categorize a face as a face, discriminate faces from objects and from a complex scene background, even at brief presentations (100 msec; Schiltz et al., in press). Her gender and expression performances are relatively well preserved, although slightly below normal range (Rossion et al., 2003). This is in stark contrast with her inability to recognize previously seen or familiar faces and to match unfamiliar faces (see Methods section). In sum, she is unable to derive an individual representation of a face that is both selective and invariant (robust) enough so that it can be discriminated from other faces and be associated with the same or other views of the same face (Rossion et al., 2003).

Below figure shows how an average and a prosopagnosic patient examine a face.

The brain area that is critically involved in facial recognition is quite large and may have further specializations within it. Brain imaging and complementing lesion studies show that when humans view an unfamiliar face, there is an increase in cerebral blood flow, and therefore activity, in a specific part of the fusiform gyrus, located more towards its back end. This zone becomes active even when humans recognize faces in a degraded visual input. By contrast when people recognize familiar face, the increase activity occurs not only in the fusiform gyrus but also in the frontal lobes. One possible conclusion to draw from these observations is that there is a further functional specialization for face perception, with posterior parts of the fusiform gyrus being devoted to the processing of signals related to faces, and to their recognition as faces, and with another subdivision of the fusiform gyrus being responsible, in association with the frontal lobes, for the recognition of the familiarity of faces so processed.

Because prosopagnosic patients are unable to recognize faces, it is no use asking a patient with a lesion in the fusiform gyrus to admire the aesthetics of portrait painting; a whole function of portrait painting is lost to such individuals. But there is a significant difference in the symptomatology of at least some types of prosopagnosia as in achromatopsia in which recognition of familiar face is impaired. Achromatopsic patient is often incapable of remembering colors or imagining what they look like, memory itself is not as badly effected here. A patient with prosopagnosia sad that he could close his eyes and visualize his wife and kids.

As outlined above, many know that they are looking at a face, although they can not recognize its identity, even if it’s their own face. The brain is unable to fit the present visual perception of a face to the specific memory record of the brain. Two syndromes may differ in their neurological basis, like achromatopsia, is a syndrome of high specificity, leading to the loss of one aesthetic sense without necessarily involving others.

http://www.achromat.org/what_is_achromatopsia.html (What is achromatopsia)

Prosopagnosic patients, those who have lost the ability to recognize familiar faces, have not necessarily lost the ability to recognize the expression on a face whose identity they are no longer able to recognize. They might, for example, have no knowledge of who a painting portrays but can tell whether the face shows characteristics of a happy or sad person. This syndrome is called vultanopsia – it is only when the lesion in the fusiform gyrus extends more anteriorly and involves other structures that patients lose the ability to recognize both the face and its expression. The large and complex almond-shaped nucleus known as the amygdala involves in affective states, and most especially fear. Monkeys without an amygdale are devoid of all sense of fear. A patient with amygdala is able recognize a face but is not able to see the fear on such face.

LEAD Technologies Inc. V1.01

http://www.sfn.org/content/Publications/BrainBriefings/fear.html (fear and the amygdala)

Correspondingly, a patient with a lesion restricted to the amygdale is able to recognize a face perfectly well but maybe specifically unable to recognize fear on such a face.

The inability to recognize expressions on faces, following lesions in the fusiform gyrus, has not been studied in the detail that one would like. The discovery itself is new and unexpected. We know only that patients with extensive, and anteriorly placed, lesions are not only incapacitated in the recognition of familiar faces but also in recognizing the broad categories of expression on a face, such as whether the face is happy or sad one. The expression on Franz Hals ‘Laughing Cavalier would, I imagine, be totally inaccessible to such a patient and the aesthetics attached to it would therefore be lost on him as well.

Franz Hals ‘Laughing Cavalier

There can also be the portrayal of fear in a portrait. Once again, the fact that we can portray fear implies that there is a certain ensemble of features that conveys fear because there is a certain neural organization that specifically recognizes fear in such an ensemble of features. It turns out that that specific organization may not be widely distributed in the visual brain but may be specific to the amygdale. The efforts of an artist to convey fear as an expression on a face would be totally lost on patients with amygdale damage. But there are even more subtle and wonderful expressions, which may engage a great deal more in the brain and about which we know little.

http://splweb.bwh.harvard.edu:8000/pages/papers/onitsuka/Onitsuka_Fusiform_2003.pdf

Without healthy functioning of the area that is shown in colors above as shown above, there can be no aesthetic of portrait painting related to the recognition of familiar faces, or just of faces. An artist would find it difficult, maybe impossible, to indulge in such an art if he had a lesion in the relevant area. Matisse, who had a great admiration for portrait painting, relates that he had a remarkable memory for faces when for those he had only seen once.

Matisse, Portrait of Madame Matisse, 1905

http://www.artchive.com/artchive/M/matisse.html

Portrait painting has many functions, all of them related directly to the need of brain to acquire knowledge. Brain seize knowledge of the characteristics of a person from portrait, through its record of past experiences, associates certain features with certain mental states and psychological traits.

In the days before photography, they were commonly used for resembling men and women of wealth and power with what their loved ones or future spouses looked like. It is not surprising that people used portrait to acquaint others with themselves; the easiest way of recognizing someone is through the face because it carries important characteristic features of the person than other parts of the body.

There is also platonic ideal with regard to the portrait of an individual. Schopenhauer says that to obtain knowledge about an object not as particular thing but as Platonic Ideal is achieved through the enduring form of this whole species of thing. Thus a great portrait should be a true likeness of an individual, no matter how that individual dressed or what angle he is captured from, to enable the brain to recognize it as Schopenhauer said it should be “the ideal of the individual”.

“The arts whose aim is the representation of the Idea of man, have as their problem, not only beauty, the character of the species, but also the character of the individual, which is called, par excellence, character. But this is only the case in so far as this character is to be regarded, not as something accidental and quite peculiar to the man as a single individual, but as a side of the Idea of humanity which is specially apparent in this individual, and the representation of which is therefore of assistance in revealing this Idea.”

The first thing when you look at a portrait, or the appearance of a face on a canvas, is that it commonly dominates, even if it does not constitute the predominant parts in terms of size or the light reflected from it. In Fantin-Latour’s Self Portrait, the intensity of light reflected from the color is much greater than that reflected from the face. It is obscured.

http://www.wetcanvas.com/Museum/Artists/f/Henri_Fantin-Latour/

And yet the face and its expression constitute the dominant elements. A far greater intensity of light reflected from color is notable by the by the Dutch painters, including Rembrant and his followers. All these works the face itself obscured in terms of the amount of light reflected from it, that is the dominant perceptual feature. Apparently, the brain is much more interested in focusing and concentrating on the face. Rest of the painting is almost non-existent- the face can survive on its own.

http://www.rembrandthuis.nl/cms_pages/index_main.html

Rembrant “ Self Portrait”

Rembrant “Portrait of the artist at his easel” 1660

Rembrant “ Portrait of a Lady with an Ostrich-Feather Fan” 1660

Titian portrait shown below said to be himself, shows a man recognizable at a glance as being somewhat remote and disdainful. Titian here uses the device of the twisted view, apparently then common in Italy, to enhance the effects of self-assuredness, his subject is looking at us with his eyes only, his head being only partially turned in our direction. Titian managed to capture on canvas an essential feature which gives immediate knowledge about that person. The portrait stands as a great portrait not because it is a likeness of Titian or indeed of any other individual but because it has captured the essential feature of haughtiness and arrogance in the brain’s record, the Platonic Ideal or the Hegelian Concept that, transposed to any face, will convey the same psychological portrait. It does not only convey information about that particular person but about all persons with similar features. The features as depicted are constant ones, always indicating a certain type of personality. It is in the classical sense idealization; in the neurological sense it distills the essential features, has elements of constancy within it.

Titian, Portrait of a man (National Gallery, London)

http://www.artchive.com/artchive/T/titian.html

That means that if the same device were used on another face, the impression of haughtiness and disdain will obtain. The expression and the psychological characteristics that it conveys are no longer tied to an individual face. Other devices can convey other expressions and these too, are not tied to an individual face but can be used on the portraits of many different individuals to convey the same psychological portrait. Two individuals who are totally unlike can be portrayed in a way that they are seen to share many psychological characteristics in common. Because subtle changes in facial expression can give different impressions, and different moods and nuances, these variations in portrait painting can also lead to subtle changes in perceptual effects.

In his portrait of Venetian Doge, Leonardo Loredan, Giovanni Bellini Managed, through subtle manipulations of the features of the two sides of the face and somewhat less subtle manipulation of the light falling on the two sides, to convey two different impressions at once; two the left, the use of a fixed gaze gives the impression of a rigid and severe person while to the right, the use of shadows and slightly more benign gaze, together with a hint of a fatherly smile, conveys the impression of a slightly more approachable person.

Giovanni Bellini, Doge Leonardo Loredan (National Gallery, London)

http://www.artchive.com/artchive/B/bellini.html

From more recent times, Picasso gives his Cubist portrait of Wilhelm Uhde a somewhat intellectual and austere look by giving the left eye a fixed glare, tightening the lips and exaggerating the furrow above the mouth.

Pablo Picasso, Portrait of Wilhelm Uhde (Private collection, photographer: Bob Kolbrener )

http://www.picasso.fr/anglais/

One masterly portrait, Girl with Pearl Earring by Jan Vermeer, possibly that of his daughter, is one which the viewer is immediately invited in. It initiates a visual dialogue with the viewer. But the portrait is also, like his other works, a masterpiece of ambiguity in the neurological sense.

http://www.artchive.com/artchive/V/vermeer.html (Jan Vermeer)

The expression on her face is at once inviting and resentful, erotically charged and demanding but also distant, somewhat sad and yet joyful, anticipating a move and yet resistant, submissive and yet dominant. From the neurological portrait of prosopagnosia as a syndrome that I have given above, one can probably make the following deductions about the perception of Vermeer’s portrait; first, that patients with lesions in the posterior part of the fusiform gyrus would not be able to see tha face at all, next that patients with lesions in the more anterior part would be able to see tha face but not recognize whose it is, and lastly patients with even more anteriorly placed lesions would be incapable of distinguishing the expression on her face. But we have little knowledge of what brain areas are involved in the powerful subjective feelings that the painting arouses, or how these brain areas interact to give us an overall impression of the painting. We are therefore still ignorant of much about the workings of the visual brain, and above all of the neurological basis of beauty. But the considerably achievements that allow us to pinpoint with an unimaginable accuracy the brain areas would not exist without all the beauty of portrait painting.

FACE RECOGNITION

Within the past decade, many systems and techniques have been developed in face recognition. Many systems have achieved recognition rates in excess of 90% accuracy under controlled conditions. In field settings, face images are subject to a wide range of variation that includes viewing, illumination, occlusion, facial expression, time delay between acquisition of gallery and probe images, and individual differences. The scalability of face recognition systems to such factors is not well understood.

There are six factors unrelated to person identity that modify face image appearance;

Viewing angle. The face has a 3D shape. As the camera pose changes, the appearance of the face can change due to projective deformation, which leads to stretching and foreshortening of different part of face, and self-occlusion and dis-occlusion of parts of the face. If we have seen faces only from one viewing angle, in general it is difficult to recognize them from disparate angles.

Pose variation still presents a challenge for face recognition. Frontal training images have better generalizability to novel poses than do non-frontal training images.

Figure 1: Pictures of the CMU 3D room setup. 10 of the 13 cameras are indicated in (a). (b) shows 17 of the 21 flash locations.

Figure 2: Pose variation in the PIE database [27]. The pose varies from full left profile (c34) to full frontal (c27) and on to full right profile (c22). The 9 cameras in the horizontal sweep are each separated by about 22:5Æ. The 4 other cameras include 1 above (c09) and 1 below (c07) the central camera, and 2 in the corners of the room (c25 and c31), typical locations for surveillance cameras.

Illumination: Lighting changes within and between days indoor and outdoor environments. Due to the 3D shape of the face, direct lighting source can caste strong shadows and shading that accentuate or diminish certain facial features.

Face recognition systems have difficulties in extreme illumination conditions in which significant parts of the face are invisible. Furthermore, it can become particularly difficult when illumination is coupled with pose variation.

Figure 3: Illumination variation in the PIE database. The figure shows twelve flash conditions

across three head poses.

Expression: The face is a non-rigid object. Facial expression of emotion and paralinguistic communication along with speech acts can and do produce large variation in facial appearance.

Deformation of the mouth and occlusion of the eyes by eye narrowing and closing present a problem for the face recognition.

Figure 4: Cohn-Kanade AU-Coded Facial Expression database. Examples of emotion-specified expressions from image sequences.

Figure 5: AR database. The conditions are: (1) neutral, (2) smile, (3) anger, (4) scream, (5) left

light on, (6) right light on, (7) both lights on, (8) sun glasses, (9) sun glasses/left light (10) sun

glasses/right light, (11) scarf, (12) scarf/left light, (13) scarf/right light

4. Occlusion: The face may be occluded by other objects in the scene or by sunglasses and other paraphernalia. Occlusion may be unintentional or intentional. Under some conditions subjects may be motivated to thwart recognition efforts by covering portions of their face.

The performance of the face recognition algorithms under occlusion is in general poor.

5. Time delay: Faces change over time. There are changes in hair style, makeup, muscle tension and appearance of the skin, presence or absence of facial hair, glasses, or facial jewelry, and over longer periods effects related to aging.

Time delay between acquisition of gallery and probe images can cause degradation in face recognition performance.

6. Individual factors: Algorithms may be more or less sensitive for men or women or members of different ethnic groups. Females might be harder to recognize because of greater use and day-to-day variation in makeup or in structural facial features. Male and female faces differ in both local features and in shape. Men’s faces on average have thicker eyebrows and greater texture in the beard region. In women’s faces, the distance between the eyes and brows is greater, the protuberance of the nose smaller, and the chin narrower than in men. People readily distinguish male from female faces using these and other differences (e.g., hair style), and connectionist modeling has yielded similar results.

In two databases (AR and FERET) the recognition rate for female subjects is higher than for males across a range of perturbations. One hypothesis is that women invest more effort into modifying their facial appearance, by use of cosmetics, for instance, which leads to greater differentiation among women than men. Alternatively, algorithms may simply be more sensitive to structural differences between the faces of women and men. The finding that algorithms are more sensitive to women’s faces suggests that there may be other individual differences related to algorithm performance. Algorithms may, for instance, prove more accurate for some ethnic groups or ages than others.

Four baseline face recognition algorithms have been developed. They are:

§ A standard PCA, or Eigenfaces, algorithm

§ A combination PCA and LDA algorithm based upon the University of Maryland algorithm in the FERET tests.

§ A Bayesian Intrapersonal/Extrapersoanl Image Diffference Classifier based upon the MIT algorithm in the FERET tests.

§ An Elastic Bunch Graph Matching Algorithm that uses localized landmark features represented by Gabor jets. This algorithm is based upon the USC algorithm in the FERET tests.

http://dmoz.org/Computers/Security/Biometrics/Face_Recognition/

EASY AND DIFFICULT SCENARIOS IN FACE RECOGNITION

Developments in Face Recognition Technology

§ Face Recognition Demos:

o Photobook/Eigenfaces

o Automatic Face Processor

o The FERET Face Database
http://vismod.media.mit.edu/vismod/demos/facerec/

§ Twins crack face recognition puzzle CNN

invented face recognition application btw twins after Sept 11

http://www.cnn.com/2003/TECH/ptech/03/10/israel.twins.reut/

§ The Tampa Police Department was testing out a new technology, called FaceIt, that allows snapshots of faces from the crowd to be compared to a database of criminal mugshots.
http://computer.howstuffworks.com/facial-recognition.htm/printable

§ Can face recognition keep airports safe?

As U.S. airports begin installing face-recognition systems to thwart terrorism in the wake of the Sept. 11 attacks, civil rights activists are rushing to decry the technology as ineffective and invasive.

http://news.com.com/2100-1023-275313.html

§ The Facial Recognition Technology (FERET) Database

The FERET program ran from 1993 through 1997. Sponsored by the Department of Defense's Counterdrug Technology Development Program through the Defense Advanced Research Products Agency (DARPA), its primary mission was to develop automatic face recognition capabilities that could be employed to assist security, intelligence and law enforcement personnel in the performance of their duties.

http://www.itl.nist.gov/iad/humanid/feret/feret_master.html

§ Bayesian recognition framework in which a model of the whole face is enhanced by models of facial feature positions and appearances.

Face recognition and facial expression recognition are carried out using maximum likelihood decisions. The algorithm finds the model and facial expression that maximizes the likelihood of a test image. In this framework, facial appearance matching is improved by facial expression matching. Also, changes in facial features due to expressions are used together with facial deformation patterns to jointly perform expression recognition.

http://www.bioforensics.com/kruglaw/f_facial.htm

§ Department of Homeland Security Adopts Facial Recognition Standard

The U.S. Department of Homeland Security announced today the adoption of its first biometric facial recognition standard. The standard is designed to be consistent with international standards for biometrics used in such applications as travel documents. This standard will also be used to specify definitions of photographic properties and digital image attributes, and as a standards format for relevant applications, including human examination and computer automated face recognition.

http://www.dhs.gov/dhspublic/display?content=4080

§ THREE APPROACHES FOR FACE RECOGNITION.

The face recognition problem is studied. Face normalization procedure is presented. Methods of face recognition such as geometric approach, elastic matching and neural networks are presented.

http://handysolution.com/facerec.htm

§ Glasgow Face Recognition Group
Average faces How to recognise faces over a huge range of images
The IAC model A model of cognitive aspects of face recognition, including a web based IAC model which will run in your browser)
CCTV Identification of faces from video (human and automatic)
http://staff.psy.gla.ac.uk/~mike/facerec.html

§ CNN.com - Facial-recognition tech has people pegged

Visionics' FaceIt software measures a face according to its peaks and valleys -- such as the tip of the nose, the depth of the eye sockets --which are known as nodal points. "While a human face has 80 nodal points," says Zelazney, "we require only 14 to 22 to do the recognition. http://archives.cnn.com/2001/TECH/ptech/07/17/face.time.idg/

§ Cellphones learn to recognize their owners' faces

Oki Electric Industry Co. Ltd. this week began marketing a technology that inexpensively adds face recognition to camera-equipped cellphones.

http://www.deviceforge.com/news/NS2876211743.html

§ The Mathematics of Face Recognition

On March 10 The New York Times reported that an Internet security consultant, doubting that the "disheveled, dazed-looking man" arrested as Khalid Shaikh Mohammed was the man shown on the FBI's most-wanted posters, fired off messages to producers of face recognition systems, asking them to compare the arrest and poster images. http://www.siam.org/siamnews/04-03/face.htm

§ Face Perception and Recognition Laboratories

Alice J. O'Toole, Ph.D.

The University of Texas at Dallas

We used high-level configural aftereffects induced by adaptation to realistic faces to investigate visual representations underlying complex pattern perception. We found that exposure to an individual face for a few seconds generated a significant and precise bias in the subsequent perception of face identity.
http://www.utdallas.edu/~otoole/face_try.html

§ Face-recognition software may aid search for movie Buddha
Indian financial newspaper Business Standard reports that Buddha Films is talking to Google and other tech firms about using facial recognition software to find an actor to play Buddha.

http://www.boingboing.net/2006/01/08/facerecognition_soft.html

§ World's First Face Recognition Biometric for Mobile Phones

http://www.japancorp.net/Article.Asp?Art_ID=9494

§ Face Recognition Using Component-Based

SVM Classification and Morphable Models

We present a novel approach to pose and illumination invariant face recognition that combines two recent advances in the computer vision field: component-based recognition and 3D morphable models.

http://www.mpi-sb.mpg.de/~blanz/publications/HuangBlanzHeisele02.pdf

§ AUTOMATIC FACE RECOGNITION OF VIDEO SEQUENCES USING SELFEIGENFACES

The objective of this work is to provide an efficient face recognition scheme useful for video indexing applications

http://gpstsc.upc.es/GTAV/Torres/Publications/ISIC00_Torres_Lorente_Vila.pdf

§ Face Recognition Based on Frontal Views Generated from Non-Frontal Images

This paper presents a method for face recognition across large changes in viewpoint. Our method is based on a Morphable Model of 3D faces that represents face-specific information extracted from a dataset of 3D scans.

http://www.mpi-sb.mpg.de/~blanz/publications/cvpr05_blanz.pdf

§ KEN Project: Real-World Face Recognition

THE Institute for Scientific Computing Research at the Laboratory recently developed the real-time, face-recognition system KEN.

http://www.llnl.gov/etr/pdfs/10_94.3.pdf

Research

Zeki Semir “Inner Vision”

http://www.face-rec.org/

http://www.frvt.org/

http://www.cs.colostate.edu/evalfacerec/

http://www.epic.org/privacy/facerecognition/

http://www.cs.cmu.edu/afs/cs.cmu.edu/user/mitchell/ftp/faces.html

http://www.cs.cmu.edu/afs/cs.cmu.edu/user/avrim/www/ML94/face_homework.html