Li S.Z., Jain A.K. (eds.) Encyclopedia of Biometrics

Подождите немного. Документ загружается.

G, and B components are highly correlated. The lumi-

nance of a given RGB pixel is a linear combination of

the R, G, and B values. Therefore, changing the lumi-

nance of a given skin patch affects all the R, G, and B

components. In other words, the location of a given

skin patch in the RGB color cube will change based on

the intensity of the illumination under which such

patch was imaged! This results in a very stretched

skin color cluster in the RGB color cube. This can be

noticed in the ﬁrst row of Fig. 1 where skin patches

from images of Asian people taken at arbitrary random

illumination are plotted in the RGB space. The skin

color cluster is extended in the space to reﬂect the

different illumination intensities in the patches. Simi-

larly, the skin color clusters for patches from different

races will be located at different locations in the RGB

color space. This can be seen in the ﬁrst row of Fig. 2.

Despite these fundamental limitations, RGB is exten-

sively used in skin detection literature because of its

simplicity. For exampl e, RGB is used by Rehg and

Jones [8] and yields quite a satisfying performance.

TV Color Spaces and Skin Detection: A different

class of color spaces is the orthogonal one used in TV

tansmission. This includes YUV, YIQ, and YC

. YIQ

is used in NTSC TV broadcasting while YC

is used

in JPEG image compression and MPEG video com-

pression. One advantage of using these color spaces is

that most video media are already encoded using these

color spaces. Transforming from RGB into any of these

spaces is a straight forward linear transformation [5].

All these color spaces separate the illumination channel

(Y) from two orthogo nal chrominance channels (UV,

IQ, C

). Therefore, unlike RGB, the location of the

skin color in the chrominance channels will not be

Skin Detection. Figure 1 Density plots of Asian skin in different color spaces.

1220

Skin Detection

affected by changing the intensi ty of the illumina-

tion. In the chrominance channels the skin color is

typically located as a compact cluster with an elliptical

shape. This can be seen in Figs.1d–f. This facilitates

building skin detectors that are invariant to illu-

mination intensity and that use simple classiﬁers. The

density of the skin color over the chrominance chan-

nels can be easily approximated using a multivariate

Gaussian distribution. Moreover, the skin colors of

different races almost colocate in the chrominance

channels. This can be seen in the second and third

rows of Fig. 2. Therefore, using such color spaces

results in skin detectors which are invariant to

human race. The simplicity of the transformation

and the invariant properties made such spaces widely

used in skin detection applications [1, 2, 9–11, 14].

Skin Detection. Figure 2 Density plots of Asian, African and Caucasian skin in different color spaces.

Skin Detection

1221

Perceptual Color Spaces and Skin Detection: Percep-

tual color spaces, such as HSI , HSV∕HSB, and HSL

(HLS), have also been popular in skin detection. These

color spaces separate three components: the hue (H),

the saturation (S), and the brightness (I,V or L). Essen-

tially, HSV-type color spaces are deformations of the

RGB color cube and they can be mapped from the

RGB space v ia a nonlinear transformation. One of

the advantages of these color spaces in skin detection

is that they allow users to intuitively specify the bound-

ary of the skin color class in terms of the hue and

saturation. As I, V or L give the brightness information,

they are often dropped to reduce illumination depen-

dency of skin color. These spaces have been used

by Shin et al. [11] and Albiol et al. [2].

Colorimetric Color Spaces and Skin Detection:

Separating the chromaticity from the brightness is

also achieved in Colorimetric color spaces, such as

CIE-XYZ, CIE-xy, CIE-Lab deﬁned by the Interna-

tional Commission on Illumination (Commission

Internationale d’E

clairage – CIE). CIE-XYZ color

space is one of the ﬁrst mathematically deﬁned color

spaces (deﬁned in 1920s). It is based on extensive

measurements of human visual perception, and serves

as a foundation of many other colorimetric spaces.

CIE-XYZ can be achieved through a linear coordinate

transformation of the RGB color space. The Y compo-

nent corresponds to the lightness of the color (the

luminance). The chromaticity values (x , y) can be

achieved by central projection into the plane XþYþ

Z¼1 and then projecting into the XY plane. For details

see [5]. The result is the well-known horse-shaped

CIE-xy chromaticity diagram deﬁning the hue and

saturation of any color. One of the disadvantages of

the XYZ and the xy color spaces is that the color

differences are not perceived equally in di fferent

regions of the color space. In contrast, the CIE

-Lab

separates a luminance variable L from two perceptually

uniform chromaticity variables (a, b). Fig.1h shows

the skin color density for Asian skin in the a,b chro-

maticity space. Figure 2 (last row) shows the skin

color density for different races in the a,b space. Despite

the many advantages of such color spaces, they are

rarely used in skin detection. This is mainly because the

transformation from RGB is more computationally

expensive than other spaces. CIE-XYZ color space

was used by Shin et al. [11] in comparison with other

color spaces. The chrominance xy plane was used by

Lee and Yoo [9].

Skin Classifiers

Avariety of classiﬁcation techniques has been used in the

literature for the task of skin classiﬁcation. A skin classi-

ﬁer is a one-class classiﬁer that deﬁnes a decision bound-

ary of the skin color class in a feature space. The feature

space in the context of skin detection is simply the color

space chosen. Any pixel which color falls inside the skin

color class boundary is labeled as skin. Therefore, the

choice of the skin classiﬁer is directly induced by the

shape of the skin class in the color space chosen by a skin

detector. The more compact and regularly shaped the

skin color class, the more simple the classiﬁer.

The simplest way to decide whether a pixel is

skin color or not is to explicitly deﬁne a boundary.

Brand and Mason [4] constructed a simple one-di-

mensional skin classiﬁer: a pixel is labeled as a skin if

the ratio between its R and G channels is between a lower

and an upper bound. They also experimented with one-

dimensional threshold on IQ plane of YIQ space where

the ‘‘I’’ value is used for thresholding. Other methods

explicitly deﬁne the skin color class boundary in a two-

dimensional color space using elliptical boundary models

[9]. The parameters of the elliptical boundary can be

estimated from the skin database at the raining phase.

Baysian Approach for Skin Detection: Skin classiﬁca-

tion can be deﬁned probabilistically as: given a pixel with

color c what is the probability of it being skin pixel

P(skinjc). Once this probability is computed, the pixel

is labeled as a skin pixel if such probability is larger

than a threshold and nonskin otherwise. Obviously

such probabilities cannot be computed for every pos-

sible color (e.g., in 24 bit RGB, there are 256

colors).

Fortunately, using Bayes rule, this can be rewritten as

PðskinjcÞ¼

Pðcjskin ÞPðskinÞ

Pðcjskin ÞPðskinÞþPðcjnotskinÞPðnotskinÞ

Bayes rule deﬁnes the posterior probability of a pixel

being skin given its color (P(skinjc)) in terms of the

likelihood of observing such color given the skin class

(P(cjskin)) and the prior probability of the skin class

P(skin). The prior probability measures our guess about

a random pixel being a skin without observing its color.

The denominator in the Bayes rule is the total proba-

bility of observing the color c, a factor that does not

affect the decision whether a pixel ought to be labeled as

skin or nonskin. Given Bayes rule, the skin classi-

ﬁcation reduces to computing the likelihood term, i.e.,

P(cjskin). Given a database of skin-colored pixels, the

1222

Skin Detection

probability density function (pdf) of P(cjskin) can be

estimated. Several approaches have been introduced

to compute this pdf including the use of histograms

[8], the use of a single Gaussian model, or a Mixture of

Gaussians model [12] to approximate such pdf.

The skin classiﬁer can also be posed as a two-class

problem. From Bayes rule, this results in computing the

likelihood ratio of observing a given color given a skin

class versus a nonskin class, i.e., P(cjskin)∕P(cjnotskin).

Such a ratio can then be thresholded to decide whether

a pixel is a skin or nonskin pixel. Besides modeling the

likelihood of an observed color given the skin class, the

complementary class needs to be models. That is,

modeling the probability density function of nonskin

pixels P(cjnotskin). Rehg and Jones [8] approximated

such pdfs using 3D histograms in the RGB space based

on a large database of skin and nonskin images.

Skin Detection Applications and

Examples

Human face localization and detection is the ﬁrst step

in obtaining face biometrics. Skin color is a distin-

guishing feature of human faces. In a controlled back-

ground environment, skin detection can be sufﬁcient

to locate faces in images. As color processing is much

faster than processing other facial features, it can be

used as a preliminary process for other face detection

techniques [10]. Skin detec tion has also been used to

locate body limbs, such as hands, as a part of hand

segmentation and tracking systems, e.g., [7 ].

Forsyth and Fleck [6] demonstrated that skin ﬁlter

can be used as part of the detection process of images

with naked or scantily dressed people. Their technique

has three steps. First, a skin ﬁlter, based on color and

texture, was used to select images with large areas of

skin-colored pixels. Then, the output is fed into a

geometric ﬁlter which identiﬁes the skin-colored regions

with cylindrical shapes. Those skin-colored cylinders

are grouped into possible human limbs and connected

groups of limbs. Images containing sufﬁciently large

skin-colored groups of possible limbs are then reported

as containing naked people.

Zheng et al. [14] presented an adaptive skin detec-

tor for detecting naked pictures on the internet. Their

technique applies a face detector on the picture ﬁrst to

ﬁnd the skin color. They argued that as skin color highly

depends on illumination and the race of the person, it is

more appropriate to get the skin color from the face of

the person in the image. Using the skin color and the

property of the texture from the detected face region,

the rest of skin pixels in the image can be detected.

Skin Detection Performance

Regardless of the choice of the color space and the

classiﬁcation method, most published research on

skin detection reports about 95% true detection while

the false detection rates varies from 15 to 30%.

Summary

Skin detection in color images and videos is a very

efﬁcient way to locate skin-colored pixels, which might

indicate the existence of human faces and hands. How-

ever, many objects in the real world have skin-tone

colors, such as some kinds of leather, sand, wood, fur,

etc., which might be mistakenly detected by a skin

detector. Therefore, skin detection can be very useful

in ﬁnding human faces and hands in controlled envir-

onments where the background is guaranteed not to

contain skin-tone colors. Since skin detection depends

on locating skin-colored pixels, its use is limited to color

images, i.e., it is not useful with gray-scale, infrared, or

other types of image modalities that do not contain

color information. There has been extensive research

on ﬁnding human faces in images and videos using

other cues such as ﬁnding local facial features or ﬁnding

holistic facial templates [13]. Skin detection can also be

used as an efﬁcient preprocessing ﬁlter to ﬁnd poten-

tial skin regions in color images prior to applying more

computationally expensive face or hand detectors.

Related Entries

▶ Face Localization

▶ Face Recognition, Over view

▶ Hand Recognition

References

1. Abdel-Mottaleb, M., Elgammal, A.: Face detection in complex

environments from color lmages. In: Proceedings of the Interna-

tional Conference on Image Processing (ICIP), pp. 622–626 (1999)

Skin Detection

1223

2. Albiol, A., Torres, L., Delp, E.: Optimum color spaces for skin

detection. In: Proceedings of the International Conference on

Image Processing (ICIP), pp. I: 122–124 (2001)

3. Bernhard Fink, K.G., Matts, P.J., Visible skin color distri-

bution plays a role in the perception of age, attractiveness,

and health in female faces. Evol. Hum. Behav. 27(6), 433–442

(2006)

4. Brand, J., Mason, J.: A comparative assessment of three

approaches to pixellevel human skin-detection 1, 1056–1059

(2000)

5. Burger, W., Burge, M.: Digital Image Processing, an Algorithmic

Introduction Using Java. Springer, Berlin (2008)

6. Fleck, M.M., Forsyth, D.A., Bregler, C.: Finding naked people.

In: Proceedings of the European Conference on Computer

Vision (ECCV), pp. 593–602 (1996)

7. Imagawa, K., Lu, S., Igi, S.: Color-based hands tracking system

for sign language recognition. In: FG ’98: Proceedings of the

Third International Conference on Face and Gesture Recogni-

tion, p. 462. IEEE Computer Society, Washington, DC, USA

(1998)

8. Jones, M.J., Rehg, J.M.: Statistical color models with appli-

cation to skin detection. Int. J. Comput. Vision (IJCV)

46(1), 81–96 (2002). URL citeseer.ist.psu.edu/jones99statistical.

html

9. Lee, Y., Yoo, S.I.: An elliptical boundary model for skin color

detection. In: Proceedings of the International Conference on

Imaging Science, Systems, and Technology (2002)

10. Senior, A., Hsu, R.L., Mottaleb, M.A., Jain, A.K.: Face detection

in color images. IEEE Trans. Pattern Analy. Mach. Intell.

(PAMI) 24(5), 696–706 (2002). DOI http://dx.doi.org/10.1109/

34.1000242

11. Shin, M.C., Chang, K.I., Tsap, L.V.: Does colorspace transforma-

tion make any difference on skin detection? In: WACV ’02:

Proceedings of the Sixth IEEE Workshop on Applications of

Computer Vision, p. 275. IEEE Computer Society, Washington,

DC, USA (2002)

12. Yang, M., Ahuja, N.: Gaussian mixture model for human skin

color and its application in image and video databases.

In: Proceedings of the SPIE: Conference on Storage and

Retrieval for Image and Video Databases (SPIE 99), vol. 3656,

pp. 458–466 (1999)

13. Yang, M., Kriegman, D., Ahuja, N.: Detecting faces in images:

A sur vey. IEEE Trans. Pattern Analy. Mach. Intell. (PAMI) 24(1),

34–58 (2002)

14. Zheng, Q.F., Zhang, M.J., Wang, W.Q.: A hybrid approach to

detect adult web images. In: PCM (2), pp. 609–616 (2004)

Skin Print

▶ Skin Texture

Skin Spectroscopy

DONG YI,WEILONG YANG,STAN Z. LI

Center for Biometrics and Security Research &

National Laboratory of Pattern Recognition, Institute

of Automation, Chinese Academy of Sciences,

Beijing, China

Synonyms

Spectral analysis of skin

Definition

Skin spectroscopy is the study of interaction between

radiation and human skin, as a function of wavelength.

Human skin has a complicated multilayered structure

and each layer is composed of different chemical sub-

stance. The determination of the interaction can be

made by measuring the reﬂectance property of skin.

Introduction

The interaction between electromagnetic radiation and

matter at various wavelengths can be used to reveal the

structure of matter. The study of this interaction is

named as spectroscopy. Since the composition of the

object has different response under different radiation,

when the surface of an object is illuminated by the

radiation with different wavelength, the reﬂectance

will vary as well.

The study of skin spectroscopy originated in the

ﬁeld of photobiology, which has tens of years history.

From the reﬂectance of skin under the radiation of

some wavelength, clinicians can obtain the composi-

tion of skin and observe the changes in skin or blood.

In [1], Anderson et al. established an effective optical

model for each layer of human skin by Kulberka-Munk

approach. The work developed a quantitative, general

model for the radiation transfer in the human skin.

Based on this model, the optical paramet ers of each

skin layer can be acquired.

Skin is a good basis for establishing the biometric

identity, because everyone has unique skin properties

in terms of color, appearance, texture, and inner struc-

ture. Skin spectroscopy provides an efﬁcient way to

1224

Skin Print

describe these properties of skin under the radiation of

different wavelengths. Measurements of skin optics

under radiations of multiple wavelengths can provide

more information than using a single wavelength,

which can enhance the performance of biometric

system. In addition to the traditional visible light im-

aging, near-infrared and thermal-infrared imaging

have been applied in the biometric system.

Spectral Bands

Skin spectroscopy is focused on the interaction be-

tween electromagnetic wave and human skin. From

long to short wavelength, electromagnetic wave can

be divided into several categories: radio wave, infrared

(IR), visible light (VIS), ultraviolet (UV), x-ray,and

gamma ray. Radio wave is usually passed around the

human body with no interaction with the human skin,

due to its long wavelength and low energy. X-ray and

gamma ray can easily penetrate human skin to damage

cells in tissue. Although ultraviolet is widely used in

photo-medicine for disease diagnosis, high-intensity

ultraviolet can cause skin suntan, burn, or even skin

cancer. For noninvasive biometric applications, VIS

and IR are two main bands that have been used practi-

cally. The spect rum of visible light is usually divided

into three channels: red, green, and blue. The infrared

portion of electromagnetic wave can be divided into

four spectral regions: near infrared (NIR), short-wave

infrared, thermal infrared (TIR), and far infrared

(FIR). Fig. 1 illustrates UV, VIS, and IR bands in

different wavelength.

Human Skin Structure

Skin is the largest organ of the human body, which

covers the whole body and protects the internal tissue

from outside damage. From one’s skin, we can access

his mood, health condition, and attractiveness. Human

skin has a complex, multilayered structure and chemi-

cal composition. Generally, from outside to inside,

human skin is composed of the following layers:



The epidermis, which is the exterior layer of skin. It

does not contain blood vessels. The main type of

cells in the epidermis are keratinocytes, melano-

cytes, Langerhans cells, and Merkels cells. Among

these cells, melanocytes are the most important

cells for skin spectroscopy, because they can syn-

thesize melanin.



The dermis, which is the layer of skin beneath the

epidermis. It contains hair follicles, sweat glands,

sebaceous glands, apocrine glands, lymphatic ves-

sels, blood vessels, and nerve endings. The blood

vessels provide nutrition for the epidermis and the

dermis. Nerve endings can provide the sense of

touch and heat.



The subcutaneous tissue, which has no evident

boundary with the dermis, can isolate body from

heat and store energy. The main constituent of

subcutaneous tissue is adipose.

Human Skin Optics

The radiation ﬁrst interacts with the surface of the

skin and then penetrates inside, but it usually cannot

reach the subcutaneous tissue. Thus, the skin sur-

face, the epidermis, and the dermi s are three main

aspects that been studied in skin spectroscopy. In the

following sections, the optical properties of interface

between air and the stratum corneum, the epidermis,

and the dermis will be described. The optics of skin

is dependent on the wavelength and the dose of

the incident light. Due to the high energy, x-ray

and gamma ray can cause chemical reactions with

the cells of the skin and the body, which is called

‘‘radiolysis’’. Here, only the spectrum from 0.25 to

3 mm (including UV, VIS and IR) is discussed ,

because the other spectra do not have much signiﬁ-

cance for biometrics.

Skin Spectroscopy. Figure 1 The electromagnetic wave radiation spectrum bands.

Skin Spectroscopy

1225

A typical process of interaction between radiation

and the skin is illustrated in Fig. 2. Under the incident

radiation, a small fraction of radiation is reﬂected by

the interface between air and the stratum corneum.

Generally, the reﬂectance is always between 4% and

7% over the entire spectrum from 0.25 to 3 mm, for all

kinds of skin (white, black, etc.). Since the surface of the

stratum corneum is not ﬂat and smooth, the reﬂectance

from skin is not specular except those reﬂected by oil.

The remaining 93–96% of the incident radiation pene-

trates into the skin and will be absorbed or scattered.

In the UV band, there are many cells and chemical

constituents in the epidermis, affecting the transmit-

tance and the remittance of the radiation. For example,

aromatic amino acids tryptophan and tyrosine are

with a minimum transmittance near 0.275 mm; nucleic

acids are with an absorption maximum near 0.26mm;

and numerous small aromatic molecules are with an

absorption maximum at 0.277 mm, and so on. The

content and distribution of melanin play an important

role in determining the optical properties of epidermis.

In the VIS band, melanin is the only pigment

affecting the transmittance of epidermis. Shorter wave-

lengths, such as blue light, are hig hly absorbed by

certain tissue components such as melanin and blood

compared with longer wavelengths. In addition, the

optical scattering increases as the wavelength gets

shorter in this spectral range. For these reasons, longer

wavelengths (red) penetrate the tissue deeper than

shorter wavelengths. This phenomenon is illustrated

in Fig. 3, in which the R channel is shown to be the

smoothest channel and B channel contains most details.

The skin color (VIS spectrum) is affected by

the type and quantity of me lanin in the epidermis.

This results in various skin colors from white to

brown and black, although the structure of human

skin is similar across different races. The absorption

of melanin decreases monotonously from the short

wavelength of 0.25 mm in the UV band (through the

VIS band wavelength) to the long wavelength of 1.1

mm in the NIR band. Beyond 1.1mm, the absorption of

melanin is negligible and both the transmittance and

remittance of skin are uncorrelated with melanin. TIR

of wavelength between 2.4 and 15 mm is often used

passively to determine the temperature of skin.

The dermis has a signiﬁcantly different structure

and composition with epidermis, which causes their

different optical properties. Maybe, because the dermis

is deeper than epidermis, fewer studies have concen-

trated on dermal optics. In the dermis, scattering holds

the dominant position. Many observations show that

the transmittance and remittance are close to 100%

across the spectrum from 0.3 to 2.4 mm, indicating

that very little radiation is absorbed there. Light of

longer wavelengths can penetrate the dermis deeper

than shorter wavelengths.

Skin Spectroscopy. Figure 2 The general model of interaction between radiation and skin.

1226

Skin Spectroscopy

Biometrics Applications

From the foregoing discussion, we can ﬁnd out two

important properties of skin spectroscopy: uniqueness

and diversity, which make the skin spectroscopy be-

come a valuable tool for the biometric technology. The

uniqueness is that everyone has his unique spectro-

scopic properties of skin, which provides a discrimina-

tive feature for the biometric systems. The diversity is

that the properties of different skin parts are diverse

under the radiation of different wavelengths, which

provides abundant features to describe skin.

Due to the nutritious texture information, face,

ﬁngerprint, and palmprint are the three main parts of

skin used in biometric system s, which are called ‘‘mo-

dality’’ in biometrics. In addition, their physical for-

mation is stable and all of them can be easily imaged by

the optical sensors with different wavelengths. Then,

the application of skin spectroscopy in the biome tric

systems is introduced in the following section.

In terms of the spectrum of radiation, VIS, NIR, and

TIR are the three common bands used widely in face

recognition. Fig. 4 shows VIS, NIR, and TIR

(thermogram) face images. Apparently, TIR face image

contains less texture information than VIS and NIR

images. The most conv entional face recognition is based

on face images captured by common VIS camera. More

recently, techniques based on NIR [2, 3] and TIR [4, 5]

images are introduced to overcome problems arising

from changes in ambient illumination. Notably, facial

thermogram is not based on skin spectroscopy, which

shows the active TIR radiation of live human skin.

In ﬁngerprint recognition, there are three common

data acquiring devices: optical, ultrasonic, and capaci-

tance. The ﬁrst one belongs to spectroscopy-based

methods. Optical imaging devices usually capture a

digital image of ﬁngerprint, using VIS light. One dis-

advantage of this type of sensor is that the imaging

capacity is affected by the skin quality of the ﬁnger.

Novel sensing techniques such as multispectral imag-

ing [7, 8] have been developed to overcome this prob-

lem. The principle of the ultrasonic device is very

similar to that of skin spectroscopy-based methods

while just using ultrasonic instead of electromagnetic

wave. Ultrasonic sensors use very high-frequency

sound waves to penetrate the epidermal layer of the

skin. Since the dermal skin layer exhibits the same

characteristic pattern of the ﬁngerprint, the reﬂected

wave measurements can be used to form an image of

the ﬁngerprint. This eliminates the need for clean,

undamaged epidermal skin and a clean sensing surface.

For palmprint biometric, a large region of palm

supplies plenty of line patterns that can be easily cap-

tured by a low-resolution imaging device. Palmprint

images are often captured by an active VIS lighting

imaging device in a semiclosed environment [9].

Skin Spectroscopy. Figure 3 Two face images in VIS and their RGB channels.

Skin Spectroscopy

1227

Strictly, the iris is not the skin, but it has a rich texture

like skin that can be used to distinguish different people.

Its complex patterns co ntain many distinctive features.

NIR wa v elengths can penetrate the iris as deep as they

penetrate the skin. Hence, NIR is used for clear and

unobtrusive imaging at a distance of up to 1m further.

Moreo v er, under active NIR light, even darkly pigmented

irises reveal rich features. Fig. 5 is a NIR iris image that

presents complex iris texture pattern.

Moreover, compared with these single-modal sys-

tems described earlier, hybrid biometric systems can

achieve better performance by fusing some modalities

or spectrum, for example, face+ﬁngerprint [10],

face+iris [11], face+palmprint [12], multispectral

NIR face [2] etc.

Summary

From the perspective of skin spectroscopy, the human

skin has its reﬂectance and absorption character-

istics as a function of wavelength of illumination.

Since the composition of skin across people have

signiﬁcant difference, skin spectroscopy has been

widely used in biometrics, in which many parts of

skin with nutritious texture (face, ﬁngerprint, palm-

print, and other modalities) are often chosen as

objects of study. For every modality, one or many

appropriate bands can be chosen according to the

skin spectroscopy. By fusing some modalities or spec-

trum, lots of hybrid biometrics have been generated.

Generally, compared with single biometric, multi-

spectral or multimodal features of skin can lead to

more effective bio metric systems. More applications

of skin spectroscopy to biometric technologies can

be foreseen.

Related Entries

▶ Face Recognition

▶ Fingerprint Recognition

▶ Multibiometrics

▶ Multi-Spectral Biometrics

▶ Near Infrared

▶ Near Infrared Face Recognition

▶ Palmprint Recognition

Skin Spectroscopy. Figure 4 VIS, NIR and TIR face images. The TIR image is from the Equinox database [6].

Skin Spectroscopy. Figure 5 A NIR iris image.

1228

Skin Spectroscopy

References

1. Anderson, R.R., Parrish, J.A.: The optics of human skin. J. Invest.

Dermatol. 77, 13–19 (1981)

2. Pan, Z.H., Healey, G., Prasad, M., Tromberg, B.: Face recognition

in hyperspectral images. IEEE Trans. Pattern Anal. Mach. Intell.

25(12), 1552–1560 (2003)

3. Li, S.Z., Chu, R., Liao, S., Zhang., L.: Illumination invariant face

recognition using near-infrared images. IEEE Trans. Pattern

Anal. Mach. Intell. 26(Special issu e on Biometrics: Progress

and Directions) (2007)

4. Selinger, A., Socolinsky, D.A.: Face recognition in the dark. In:

Proceedings of IEEE Computer Society Conference on Computer

Vision and Pattern Recognition Workshops, Washington, DC,

USA, pp. 129–129 (2004)

5. Chen, X., Flynn, P.J., Bowyer, K.W.: Infra-red and visible-light

face recognition. Comput. Vis. Image Underst. 99, 332–358

(2005)

6. Miezianko, R.: Terravic Research Infrared Database. http://www.

terravic.com/research/index.htm (Accessed June 17, 2008)

7. Rowe, R.K., Corcoran, S.P., Nixon, K.: Biometric identity

determination using skin spectroscopy. Tech. rep., http://www.

lumidigm.com (2004)

8. Jain, A.K., Flynn, P., Ross, A.A. (eds.): Handbook of Biometrics.

Springer (2008)

9. Zhang, D., Kong, W.K., You, J., Wong, M.: Online palmprint

identiﬁcation. IEEE Trans. Pattern Anal. Mach. Intell. 25(9),

1041–1050 (2003)

10. Hong, L., Jain, A.: Integrating faces and ﬁngerprints for personal

identiﬁcation. IEEE Trans. Pattern Anal. Mach. Intell

11. Wang, Y., Tan, T., Jain, A.K.: Combining face and iris for identity

veriﬁcation. In: Proceedings of International Conference on

Audio- and Video-based Biometric Person Authentication.

Guildford, UK (2003)

12. Gao, Y., M.Maggs: Feature-level fusion in personal identiﬁca-

tion. In: Proceedings of IEEE Computer Society Conference

on Computer Vision and Pattern Recognition. San Diego, CA,

USA (2005)

Skin Texture

XIANGXIN ZHU,ZHEN LEI,STAN Z. LI

Biometrics and Security Research & National

Laboratory of Pattern Recognition, Institute of

Automation, Chinese Academy of Sciences, Beijing,

China

Synonyms

Skin Print

Definition

Generaltly, skin texture is the surface texture pattern of

any part of human body with bare skin (e.g., face,

hand, and palm). In the context of biometrics, this

term commonly refers to the technologies and meth-

ods of face recognition using highly detailed facial skin

texture in high-resolution images.

Introduction

Since the success of local features in face recognition in

late 1990s, researchers have been seeking for more de-

tailed representations of human faces. Skin texture con-

tains plentiful detailed local information, and therefore

starts to attract research attention [1]. A person’s skin

texture pattern is, to some extent, a uni que physical

trait and is distinguishable from those of others, and

thus can be used for biometric identiﬁcation.

The primary difference between conventional face

recognition method s and skin texture methods is

the resolution of facial images used. Typically, in con-

ventional face recognition methods, the faces are scaled

to 30–60 or so pixels between the centers of eyes. For

skin texture methods, the inter-eye distance should be

at least 90 pixels to obtain reasonable performance.

With much higher

▶ image resolution, more sophisti-

cated face recognition algorithms could possibly be

proposed to yield better recognition accuracy.

However, with the increased facial image resolu-

tion, higher quality image capture devices, and more

computing resources are needed for acquiring and

processing the facial images, both expensive. In the

1990s, those were the main obstac les in the applica-

tions and developments of skin texture methods. Since

2000, high quality digital cameras and webcams are

becoming more affordable, and the processing speed of

personal computers is ten times faster than it was a

decade ago with much lower price. Those have cata-

lyzed the advance of skin texture technology.

Skin Texture Based Methods

The ﬁrst work to create the skin texture face recogni-

tion algorithms was pioneered by Delean Vision

around 2001 [2]. The Delean method uses a probabi-

listic image analysis method for skin texture matching.

Skin Texture

1229