(50*1/9)+ - cpe@kmutt

builderanthologyAI and Robotics

Oct 19, 2013 (3 years and 10 months ago)

88 views

CPE 631

Image Processing

and Computer Vision

2

Syllabus
-

Overview

To introduce students to the concepts of

computer vision touching on areas of


computer graphics


image processing


artificial intelligence


biological vision


neural networks


pattern recognition


robot vision

3

Syllabus
-

Application


Most applications involve us to use eyes


Good robot should be able to see


Engineering wise control environment


General environment unconstraint


Simulation
-

Driving on a highway


Millions of Research Topic

4

Syllabus
-

Grading


Programming Assignments

25%



Term Project



15%


Best
8

of
10

Quiz




60%

5

Syllabus


Term Project


The term project will be on a computer vision
topic of choice requiring
:



Proposal
:
1
-
2 pages and 5
-
10 min presentation


Project Report
:
4
-
8 pages and 20
-
25 min
presentation

6

Syllabus


Assignment

1.

Simple Thai OCR Competition
(10%)


2.

Three Dimensional Digitizer Application: Shape
from Stereograms
(15%)

7

Syllabus


Reference Books

Jain R
.
, R
.
Kasturi, and B
.
G
.
Schunck,
Machine Vision, McGraw
-
Hill
.

8

Syllabus


Topics

1.


Overview of Computer Vision


Course and Assignments Overview


Image Formation and Sensing


3
-
D Computer Graphics and Visual Realism


Digital Images: bw, grayscale, and color


2.

Binary Image Processing: Low
-
level


Image Filtering and Edge Detection


3.


Regions, Image Segmentation, Texture Segmentation


Blob Coloring


Contours and Boundary Detection


General Hough Technique and Applications

9

Syllabus


Topics

4.


3
-
D Computer Graphics Models Revisited


Optics, Shading, Curves and Surfaces


Energy Minimization and Relaxation Techniques


Bayesian Probability, The Pixel Lattice



5.


Depth & Shape from X


Texture, Shading, and Stereo


6.
Calibrations


Depth from Binocular Stereo


3
-
D Volume Rendering from cross
-
section images

10

Syllabus


Topics

7.

Dynamic Vision


Motion Field and Flow: Gradient vs. Matching
Methods



8.
Structure from Motion, Object Tracking


9.
Object Recognition Models

11

Computer Vision


Divided to 3 Levels:


Low Level


Mid Level


High Level

12

Low
-
Level


Edge Detection


Image Processing (Pre
-
Processing)

Image Processing

Image

Image

13

Mid
-
Level (Perception)


Segmentation/Grouping

Satellite Images

OCR



้ง

14

Mid
-
Level (Perception)


Finding Depth, Shape, Light, Materials


“Inverse of Computer Graphics”

Computer Graphics

CG

Image

Scence


-
shape

-
material


color, shiny, transparency,


texture, etc.

-
light

-
camera

Computer Vision

CV

Image

3D

Scence

15

High
-
Level (Use Knowledge)


Pattern Recognition


Obstacle Avoidance


Grasping


Object Recognition


Etc.

16

Image Formation


Intensity Image (with value 0..255)


0


black, 100


gray, 255
-

white


For example, an ideal white paper

200

200

200

200

200

200

200

200

200

200

200

200

200

200

200

200

200

200

200

200

17

Noise in Sensor


Scanned white paper

might not be all 200


It can be,

for example:

190

210

200

198

198

198

219

220

200

200

200

221

220

220

222

218

200

220

219

222

18

Gaussian Additive Noise


G
(m
,s
)

190

210

200

198

198

198

219

220

200

200

200

221

220

220

222

218

200

220

219

222

G(
20,5)

19

Salt/Pepper Noise


Random Noise


Black


pepper


White
-

Salt

20

To Get Rid of Noise


Gaussian Noise


Use Gaussian Filter



Salt/Pepper Noise


Use Median Filter

21

Filter

Input Image

Output Image

I

F

I’

I x

F = I’

Convolution

Filter

22

Filter

-

Example

1/4

1/4

1/4

1/4

45

54

52

47

50

50

51

49

52

47

54

47

48

45

51

52

51

50

50

52

Input Image

Output Image

2 x

2

Mean Filter

48

49

51

50

48

49

51

51

52

50

49

49

(50*1/4)+(45*1/4)+

(49*1/4)+(52*1/4)

23

Filter

-

Example

1
/
9

1/9

1/9

1/9

1/9

1/9

1/9

1/9

1
/
9

45

54

52

47

50

50

50

51

49

50

52

47

54

47

50

46

45

51

52

51

51

50

50

52

49

54

52

47

45

50

Input Image

Output Image

3 x

3

Mean Filter

50

50

50

50

50

49

50

50

51

49

48

49

Mirror

(50*1/9)+(50*1/9)+(45*1/9)+

(50*1/9)+(50*1/9)+(45*1/9)+

(49*1/9)+(49*1/9)+(52*1/9)

24

Mean Filter (Average)

1
/
9

1/9

1/9

1/9

1/9

1/9

1/9

1/9

1
/
9

3
x 3

Mean Filter

1/4

1/4

1/4

1/4

2 x

2

Mean Filter

25

Gaussian Filter

1

1

2

2

2

1

1

1

2

2

4

2

2

1

2

2

4

8

4

2

2

2

4

8

16

8

4

2

2

2

4

8

4

2

2

1

2

2

4

2

2

1

1

1

2

2

2

1

1

26

Gaussian Filter


Highest in the middle


Reducing factored by
s


s

low


reduce fast

s

桩h栠


reduce slow

27

Median Filter


No convolution


Arrange all values, then use the middle one


For example,

Use the middle


20, 21, 32, 23, 17, 19 , 20


Order by value

17

19

20

20

21

23

32


Use 20


20, 22, 32, 23, 17, 19 , 20,24


Order by value

17

19

20

20

22

23

24

32


Use 21

Average the middle

(20+22)/2 = 21

28

Median Filter
-

Example

20

20

32

21

22

22

20

23

21

100

56

24

22

24

22

70

71

22

23

20



2 x 2 Median Filter

20

20

20

20

21

22

22

23

21

22

24

24

22

22

22

22