Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 16

1

Computer
Vision
Lecture 10
HAAR-like Features
Hendrawan A.
Sekolah Tinggi Teknik Surabaya
HAAR-like Features
Digital Image Features used in
object recognition. They owe
their name to their intuitive
similarity with HAAR wavelets
and were used in the first real
time face detector.
One of the famous application
that use this features is Viola-
Jones Face Detector.
2

Sekolah Tinggi Teknik Surabaya


Viola-Jones Face Detector: Results
Viola-Jones Face Detector: Results
Viola-Jones Face Detector: Results
Viola-Jones Face Detector: Results

Paul Viola, ICCV tutorial


HAAR-like Features
Can be trained to recognize pets!

http://www.maclife.com/article/news/iphotos_faces_recognizes_cats

Paul Viola, ICCV tutorial


HAAR-like Features
The difference between pixels
sum of the white and black
areas

Sekolah Tinggi Teknik Surabaya


HAAR-like Features
The calculation of different not
just sum of the white are minus
sum of the black area, but every
area have his own weight.
For example: white area have -1
weight and black area have 2
weight. Then the calculation
become sum of white area cross
-1 plus sum of black area cross 2.
9

Sekolah Tinggi Teknik Surabaya


HAAR-like Features
Capture the face symmetry

1
0
Sekolah Tinggi Teknik Surabaya
HAAR-like Features
Can be extracted at any
location with any scale.

1
1
Sekolah Tinggi Teknik Surabaya
HAAR-like Features
Too many features!
location, scale, type
180,000+ possible features
associated with each 24 x 24
window
Not all of them are useful!
Speed-up strategy
Fast calculation of haar-features
AdaBoost
Selection of good features
1
2
Sekolah Tinggi Teknik Surabaya
Integral Image

Sum of pixel values in the blue area

Example:

2 1 2 3 4 3 2 3 5 8 12 15
3 2 1 2 2 3 5 8 11 16 22 28
4 2 1 1 1 2 9 14 18 24 31 39
Image Integral image
1
3
Sekolah Tinggi Teknik Surabaya
Integral Image
1 2 a = sum(1)
a b b = sum(1+2)
3 4 c = sum(1+3)
c d d=
sum(1+2+3+4)
Sum(4)d=+? a b c
Four-point calculation!

6-point
A, B: 2 rectangles
=> 8-point
9-point
C: 3 rectangles =>
D: 4 rectangles => 1
4
Sekolah Tinggi Teknik Surabaya
Feature Selection

f1 f2
f1 > (a threshold) => Face!
f2 (a threshold) => Not a Face!
1 if fi >
h=
0 otherwise

Sekolah Tinggi Teknik Surabaya


Feature Selection
Strong Classifier = (1h1 + 2h2) + ()+ (+>
TTthresold
hT)
1 2 3
<

Sekolah Tinggi Teknik Surabaya

You might also like