Você está na página 1de 20

Security and Privacy Concerns

Biometric Technique for

Detecting Human Faces in Images

Problem Statement

Given an image, identify one or more persons in the image, determining the locations and sizes of human faces in arbitrary (digital) image regardless of its position, orientation. Review of Design steps and implementation of skin color detection algorithm used in face detection. Comparison of skin color detection algorithms based on color model like RGB, YCbCr, HIS. Analyzing the weakness found in algorithm & implementing the algorithm using minimum facial feature and evaluating the results with other skin color detection algorithms.

What is Biometric Technique

Biometric system .

system

is

pattern-recognition

Biometric technology is a ways to secure human society and to extend knowledge about human nature and behavior. It refers to an automatic recognition of individuals based on a feature vector derived from their physiological and/or behavioral characteristics.

Biometric Parameters

Universality Uniqueness Permanence Collect ability Performance Acceptability Circumvention

Biometric System Performance Metrics

FAR (False accept rate) FRR (False reject rate) EER (Equal error rate) FTE (Failure to enroll rate) FTC (Failure to capture rate) Template capacity

Literature Survey

Face Detection Methods Color Model Skin Color Detection Algorithms

Face Detection Methods

Knowledge based Method Feature based Method Template based Method Appearance based Method

Back

Feature Based Method

Facial Feature Texture Skin color Multiple Feature

Color Model

RGB Color Model YCbCr Color Model HIS Color Model

Back

Skin Color Detection Algorithms

Skin Color Detection in RGB color space Skin Color Detection in YCbCr color space Skin Color Detection in HSI color space

Skin Color Detection in RGB color space


The skin colour at uniform daylight is given by,
(R > 95) AND (G > 40) AND (B > 20) AND (max{R, G, B} min{R, G, B} >15) AND (|R G| > 15) AND (R > G) AND (R > B) ....(1) The skin colour under flashlight or daylight is given by, (R >220) AND (G > 210) AND (B > 170) AND (|R G| 15) AND (R>B) AND (G > B) (2)

The RGB bounding rule is denoted as Rule A. Rule A: (1) OR (2)

Skin Color Detection in YCbCr color space


Conversion from RGB to YCbCr

Cr Cr Cr Cr Cr

1.5862 Cb + 20 ...(1) 0.3448 Cb + 76.2069 ................(2) -4.5652 Cb + 234.5652.(3) -1.15 Cb + 301.75.................................(4) -2.2857 Cb + 432.85.(5)

The YCbCr bounding rule is denoted as Rule A. Rule A: NOT((1) AND (2) AND (3) AND (4) AND (5))

Skin Color Detection in HSI color space


Conversion from RGB to HSI

H>45 OR I>245.(1) The HSI bounding rule is denoted as Rule A. Rule A:NOT (1)

Result of Algorithms
RGB

YCbCr

HSI

Proposed Algorithm

Skin Detection

Read all pixel value of an image Read r,g & b value of each pixel Calculate Normalize value of RGB

Calculate Upper & Lower bound of skin pixel

Conti..

Exclude White color

Calculate Hue from RGB color model

Condition for skin color is given by,

Hair Detection

Find intensity element from HSI model

Condition for Hair Detection

Challenges to Face Detection

Pose Presence or absence of structural components Facial expression Occlusion Image orientation Imaging conditions

From the data collected from various research papers and articles it is clear that the common RGB representation of color images is not suitable for characterizing skincolor. In the RGB space, the triple component (r, g, b) represents not only color but also luminance. Luminance may vary across a person's face due to the ambient lighting and is not a reliable measure in separating skin from non-skin region. It is found that it represents some non-skin region also as the skin region and hence false detection rate is quite high. In YCbCr color space, the accuracy is found higher than results from RGB color space. Experiments show that HSI color space is also good in classifying the skin color region. Skin colors of different people appear to vary over a wide range; they differ much less in color than in brightness. i.e. skin colors of different people are very close, but they differ mainly in intensities. It is good to make use of more than one color model for detecting skin color. Face detection is a challenging and interesting problem because faces are non rigid and have a high degree of variability in size, shape, color, and texture.

Você também pode gostar