Face Detection and Localisation By Hesham Ahmed Introduction Aims - - PowerPoint PPT Presentation

▶

Aug 19, 2022 517 likes •704 views

Face Detection and Localisation By Hesham Ahmed Introduction Aims Different Methods used in Current Research The Algorithm Results Demonstration Limitations and Areas for Extension Aims To automatically detect if a

SLIDE 1

Face Detection and Localisation

By Hesham Ahmed

SLIDE 2

Introduction

Aims Different Methods used in Current

Research

The Algorithm Results Demonstration Limitations and Areas for Extension

SLIDE 3

Aims

To automatically detect if a colour

image contains a human face and to locate that face within the image.

Cope with reasonable variations in

terms of race, multiple faces, size, pose, illumination and background.

SLIDE 4

Different Approaches

Edges Colour Feature Searching

SLIDE 5

The Algorithm

Skin detection through colour. Detection of eye/mouth regions through

colour, edge detection and heuristics.

Ellipse estimation using Hough Transform Weighting of Ellipse and Thresholding.

SLIDE 6

Skin Detection

Using a YCbCr colour

transformation.

Skin coloured pixels

form a tight cluster in Cb-Cr space.

Involves a lot of

experimentation to find correct thresholds.

SLIDE 7

Edge-detection

Firstly a Gaussian smoothing

mask is applied to remove spurious edges.

Sobel operator used

(horizontal, vertical and both diagonals)

Concentration of horizontal

edges used to detect eyes and mouths

Other edges used with Hough

transform to calculate ellipses.

SLIDE 8

Feature detection

Using combination of

Cb-Cr values and edge detection locate possible eye and mouth regions.

Threshold values and

fix in blocks of 8x8 pixels.

SLIDE 9

Hough Transform and Ellipse Selection

Hough Transform converts

points in Cartesian space into parametric space.

For every point that would

fit a particular ellipse, it’s ‘accumulator’ cell would be incremented.

Ellipses have 5 parameters

– extremely complex ( require 5 dimensional parameter space).

SLIDE 10

Results

Better on images suitable for biometric

applications especially Face recognition.

Images taken University College Dublin Colour

Face Database and personal collection of digital photos.

Single face images, multiple face images,

different lighting, poses, accessories, races,

rientation and occlusions.

SLIDE 11

Single face images

! " # " $ %

SLIDE 12

More Faces

! &$$$ ' ! (

SLIDE 13

Multiple faces

) ! )

SLIDE 14

Limitations

Not particularly successful at detecting faces

in large groups (small faces).

Due to effects of noise, thresholds, unclear

features and bias of Hough Transform.

SLIDE 15

Areas for future extension

Handle rotated faces by looking for

eye-candidates in several directions.

Develop greater sophistication in

confirming/rejecting face candidates

Locate features with greater accuracy.

SLIDE 16

Problems encountered and solutions

Lack of detail of implementations
Derived new and own ways of achieving the task e.g.

experimenting with thresholds, formulae and techniques.

Complexity of Hough Transform
Restricted to 2D –parameter space by estimating origin

from eye candidates and only considering of upright ellipses of specified proportions.

Also counted one in every two possible edges to improve

speed.

SLIDE 17

Demonstration…

SLIDE 18

Conclusions

Illustrated that one can still achieve good results

with a fraction of the computation and a simple algorithm.

Identified weaknesses in other existing research.
Also merged ideas and experimented with different