逄霖生 中國文化大學 電機工程學系. Outline Introduction Statistical Detection Models...

Preview:

Citation preview

逄霖生中國文化大學 電機工程學系

Outline Introduction Statistical Detection Models

Acquisition of Human Face Images Skin Detection Ocular Region Detection

Experimental Results ROC Evaluation Conclusions Future Works

IntroductionTo Detect human face by using statistical

methodsThe given image is treated as an random

variable. The colors and other features of data is treated

as the outcomes of the given random variable.The prior and posterior information can be used

to handle statistical data.The uncertainty of information reveals the

variations of data.The ROC curve statistically evaluates the

detection results.

Statistical Detection Models• Bayes’ Filter for Skin Detection

• Entropy Model for Eye Detection

• ROC (Receiver Operating Characteristic) curve for statistical evaluation of skin detection results

Statistical Detection Models• Bayes Rule

• Entropy

• ROC curve (Receiver Operating Characteristic

curve)• Statistical Evaluation of Detection Results

N

iii IpIp

12 ))()(log(

)(

)()|(

)(

)()|(

yp

xpxyp

yp

yxpyxP

Raw Image Data

Eye & Eyebrow Detection(Entropy analysis)

Skin Detection

(Bayes filter)

Face Detection

Mouth Detection(Color ratio

analysis)

PerformanceEvaluation

(ROC curve)

Color Conversion

Color Space Conversion1. RGB Primary colors (tri-stimulus values of

colors)

2. YCbCr Luminance & Chrominance

3. Gray Level s = T(r)

where “s” is an output image, “r” is an input image

B

G

R

C

C

Y

r

b

214.18786.93000.112

000.112203.74797.37

966.24553.128481.65

128

128

16

[Left] Original Image, [Right] Pre-selected Skin AreaNote that the eye & eye brow, mouth are not part of skin

Skin Detection• Applying a Bayes Filter to an image

where p(x) and p(y) are pdfs of random variables x and y, p(x|y) is the posterior probabilityp(y|x) is the prior probability.

)(

)()|(

)(

)()|(

yp

xpxyp

yp

yxpyxP

0 50 100 150 200 2500

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4p(x|y)

Pro

b. (n

orm

alize

d)

Color (x=[0..255])

Cb-SkinCr-SkinCb-NSkinCr-NSkin

)(

)()|()|(

yp

xpxypyxP

Skin DetectionBy using a Bayes filter and a thresholding

method, the skin detection result of an image is shown as follow:

Morphology

Morphology

Entropy• Entropy( 熵 )

p(Ii) is the probability for the outcome Ii

Measure the degrees of uncertainty for different outcomes from a given random event

N

iii IpIp

12 ))()(log(

....2.1, ))((log)(1

2 NrIpIpM

iirirr

N

jcjijc IpIp

12 1.2...M.c, ))((log)(

c

r

. 0,

,1 121

otherwise

tRGtif

M . ,0

,1 342

otherwise

tRBtif

M

ROC Curve• ROC Curve (Receiver Operating Characteristic

curve)

• ROC analysis provides tools to select possibly optimal models and to discard suboptimal ones independently from the cost context or the class distribution.

• ROC analysis is related in a direct and natural way to cost/benefit analysis of diagnostic decision or quality making.

• It is widely used in binary discrimination evaluation.

TPP True Positive Possibility =sensitivity

FNP False Negative Possibility

FPP False Positive Possibility =1-specificity TNP True Negative Possibility

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1ROC

TP

P(s

en

sitiv

ity)

FPP(1-specificity)

CrCb

0 50 100 150 200 2500

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4p(x|y)

Pro

b. (n

orm

aliz

ed)

Color (x=[0..255])

Cb-SkinCr-SkinCb-NSkinCr-NSkin

0 50 100 150 200 2500

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1TPP(sensitivity) & FPP(1-specificity)

TP

P &

FP

P

Color (x=[0..255])

TPP(Cr)FPP(Cr)TPP(Cb)FPP(Cb)

ROC Curve

Non-SkinArea

indeterminate Area

SkinArea

Cr FPP(cr)<0.96 0.96 FPP(cr)≦ TPP(cr)>0.9

Cb FPP(cb)>0.02 0.02 FPP(cb) 0.08≦ ≦ TPP(cb)<0.025

ConclusionsStatistical methods are able to classify and

detect human characteristics.

Using the prior information can help us to recognize the posterior situation.

The uncertainty of analyzed data gives the location of the area of eye.

ROC curve can determine the content of experimental results.

Future Works Adapted with Environmental

Variations

Hardware Acceleration

Recommended