644.50K
Category: softwaresoftware

Eye Detection in Images Introduction To Computational and biological Vision

1.

Eye Detection in Images
Introduction To Computational and biological
Vision
Lecturer : Ohad Ben Shahar
Written by : Itai Bechor
1

2.

Chapter Headings
Introduction
The Main algorithm:
Detecting the face area
Find a good candidates
Find the most probability For Eyes in The
Image
Conclusions and Results
2

3.

Introduction
Detecting Eyes has many applications:
• For Face Recognition
• May Be Use By The Police
• In Security Services
• Future Use In Computers Security For Login
Propses
3

4.

Introduction
The Eye is Quite Unique Feature in the Face
It might be easy to detect it more than other
elements in the face
The Objective is To detect the Closest Area
To the eyes or the Eyes
4

5.

The Algorithm Diagram
Detect face
Find radius that suits eye
Detect the edge
Detect the eyes
5

6.

Images I work with
Black and white images
Head Images On a Plain Background
Image resolution of 150x150 to 300x300
6

7.

Extraction of the face regions
Step 1
Input Image
M
N
Step 2
Canny Edge detector
Step 3
Calculate the left
and right bound
V(x)
x
7

8.

Face Region Extraction
8

9.

The Canny Edge Detector
I used Gaussian 5x5
convolution To smooth the
image to clean the noise
9

10.

Canny Edge Detector
Compute gradient of g(m,n) using to get:
and
And finally by threshold m:
11

11.

Hough Circle Transformation
in my program : I Find The Circles In The Image From Radius 1 to width/2.
A circle in 2d is :
The accumulator Holding the Votes For each Radius.
Largest vote (a,b)
r
(Xi,Yi)
Edge point
12

12.

Hough Circle Transformation
13

13.

Hough Circle Transformation
14

14.

Selecting the Eyes
Labeling Function That Find the best
Match Between Two Circles In The
Eyes
15

15.

Selecting the Eyes
Using the Following Methods:
1. Calculate the Distances between each two
circles .
2. The Slope Between The Two Circles.
3. The Radius similarity between two circles.
4. Large Number of circles in the same area
16

16.

Experimental Results
Good Results:
17

17.

Experimental Results
18

18.

Experimental Results
Bad Result: Hough Didn’t detect eye circles
19

19.

Experimental Results
Bad Result: Label
Function Didn’t detect
eyes.
20

20.

Conclusion
The Algorithm need to be improved
In Order To Improve it :
1. Need To Use A Eyes Database
2. There is special cameras that can detect the
eye using an effect called The bright pupil
effect .
21
English     Русский Rules