Aucun titre de diapositive

1
Image processing Introduction
Session 2005-2006
Ecole Centrale Marseille
2
European exchange
EGIM Ecole Généraliste Ingénieurs de Marseille
Lecturer microprocessor systems signal,
image processing Research CNRS
laboratory Institut Fresnel Image processing
Thierry GAIDON
BAC
Ph D
Classes prépa
Ecole ingénieurs
DEA
highschool
18 19 20 21
22 23 24
25 26 27
Européan program SOCRATES professor exchanges
John MASON Swansea
Thierry GAIDON Marseille
3

• Introduction au traitement des images
• ?Présentation générale du traitement des images
  numériques
• ?Acquisition des signaux et représentations
• ?Secteurs dapplications du traitement dimage
• ?Système visuel humain
• Introduction à la reconnaissance des formes
• ?Segmentation d'objets
• ?Description de formes (Descripteurs de Fourier,
  etc.)
• ?Introduction à la classification
• Restauration
• ?Les problèmes de la restauration dimages
• Exemples de reconstruction à partir de données
  incomplètes
• ?Méthodes de déconvolution
• ?Méthodes de réduction de bruit
• ?Méthodes bayésiennes pour la restauration
• Critère du maximum dentropie
• ?Exemple de restauration dimages
• Principales transformations en traitement
  dimages
• Introduction à la compression dimages

4
Content
1. Introductionoverview of digital signals,
image2D array, image acquisition, HVS, fields of
interest, applications 2. Mathematical
tools basic notations, some transforms (fourier,
dct, wavelets, ) 3. Image enhancement 4. Image
segmentation 5. Image compression
5
Introduction

• Overview of Digital Image Processing
• Image 2D array
• Image acquisition
• Human visual system HVS
• Fields of interest
• image enhancement, restoration, i. analysis, i.
  reconstruction, i. data compression
• Applications

6
Introduction
Why IP is important vision 70 of sensorial
stimuli for a person goal perception of the
world with vision interpretation, ...
A Picture is Worth a Thousand Words
Digital image processing refers to processing of
2D pictures by digital computers broader context
2D data
Image array of values represented by finite
number of bits
Ex of screen possibilities
7
History

• 1950 beginning
• 1970 increase of work
• spatial adventure, robots, multimedia
  applications
• computers tools

Digital image processing ? low level
Vision ? understanding of world
8
Digital signals

Sound
9
Representation
256x256 pixels
Picture element
Small square
Ex total data per image M x N x B bits
10
Letters
Pouvoir intégrateur de l œil et interprétation
du cerveau
11
Binary Images

• 1 bit for Each pixel
• 640x480 - 37.5 KB
• Dithering / Half-toning used for display

12
Look up table
Grey level intensity value
White --gt high level value Black --gt low level
value
Look up table
13
Sampling and quantization
QUANTIZE in z axis
white
SAMPLE in x and y axis pixel
black
resolution
256 levels are often used to represent images
14
Sampling and quantization
QUANTIZE in z axis
CAN converters
white
black
2 levels 1 bit
4 levels 2 bits
8 levels 3 bits
16 levels 4 bits
resolution
256 levels are often used to represent images
15
Quantization
2
8
1
64
32
16
16
Sampling
17
Sampling
16x16
32x32
64x64
128x128
256x256
512x512
18
Gray-scale Images

• 1 Byte for each pixel (0 to 255)
• 640x480 gt 300 KB of storage

19
Histogram
1 2 3 4 5 6 1 2 3
4 5 6 1 2 3 4 5 6 6
6 6 1 1 1 6 6 6 1
1 1 6 6 6 1 1 1
Repre senta tions
Amplitude 1,6
image
matrix
Stats
1/36

Number of points
frequency
Probability density function
20
Images

• Intensity
• Association of many sensors
• I(i,j) a S1(i,j) b S2(i,j) l Sn(i,j)
• Color images
• RGB (red, green, blue), YCrCb (luminance and
  chrominances) (3 images)
• Complex
• Radar images I Ireal j Iimaginary (2 images)

21
Images
White light
IR
UV
1 image 3 images of data
Optical prism
22
8 Bit Colour

• 1 Byte for each pixel
• 256 out of millions of colours possible
• Requires Colour Look-Up Table (LUTs)
• 640x480 - 307.2 KB

23
Colour images

• 3 Bytes for each pixel
• Supports 255x255x255 gt 16 Million colours
• 640x480 - 921.6 KB
• May be stored as 32 bit image
• extra byte to store special effect information

24
Standard Image Formats

• GIF Graphics Interchange Format, UNYSIS Corp.
  Compuserve, modified Lempel-Ziv Welch Algorithm
  for compression, limited to 8 bit colour images
• JPEG Joint Photographic Experts Group
• TIFF Tagged Image File Format, Developed by
  Aldus Corp. in the 1980s
• Postscript / Encapsulated Postscript Developed
  by Adobe, Typesetting / page layout language,
  text and vector / structure d graphics, images,
  Files are uncompressed, ASCII, large, Used in
  several popular programmes
• Microsoft BMP
• Macintosh PAINT and PICT ...

25
Digital images and sequences
L
L
C
C
Image
Sequence
Different formats or standards
L x C pixels by image
RGB,YUV(NTSC,PAL,SECAM)
8 bits / pixel
26
Acquisition
Real world 3D
Image digital 2D
Sensor eye, camera, 2D CCD, 1D CCD, scanner
Passive system
27
Acquisition
Reception
Emission
Active system
? wavelength
28
Basic elements
Basic elements of an image processing system
29
Acquisition
Xrays
30
Domaines d application
Image processing is the main part of many
acquisition systems, transmission systems and
information processing
31
Modelisation
Data acquisition system
output
input
50
50
100
100
150
150
200
200
250
250
50
100
150
200
250
50
100
150
200
250
image
Real scene
O h i n
noise distortions
Convolution
? 2D systems ? mathematical tools ? filtering
32
Chaîne de traitement
Formation of the image (tomography, radar,)
Compression (lossless or lossy)
TRANS- MISSION
SENSOR
Enhancement (histogram, colors,...)
VISUALIZATION
Restoration (filtering, deconvolution)
Recognition (Automatic or supervised decisions)
Segmentation (extraction of informations)
AUTOMATIC PROCESSING
33
Applications

• Remote sensing satellites, aircraft, spacecraft
• Image transmission, storage
• medical
• radar, sonar
• acoustic
• automated inspection of industrial parts

Satellites useful in tracking or earth
ressources geographical mapping prediction of
agricultural crop urban growth weather flood
and fire control ...
Thematicians ? Scientifics
34
Examples of analysis tasks
Equalization of histogram
Image enhancement
35
Idem
36
Télécommunications

• TV numérique, Internet, Téléphonie,
• Compression dimages, cryptage, tatouage

Image compressée Taux 1 256
Image pleine précision
Image compressée Taux 1 64
37
Examples
38
(No Transcript)
39
(No Transcript)
40
Other examples
41
Imagerie spatiale

• Observation du ciel (astronomie) et de la Terre
  (télédétection)
• compression, restauration, segmentation,
  reconnaissance

Image du ciel, dans la bande U.V.
Image Radar
Image SPOT (optique)
42
Vision industrielle, robotique

• Contrôle industriel, biométrie
• segmentation, reconnaissance

Images infrarouges Répartition de chaleur
43
Imagerie médicale

• Diagnostic, imagerie fonctionnelle, télémédecine,
  biologie, ...
• formation dimages, compression, restauration,
  segmentation, détection,...

Coupe radiographique dun crâne
Image échographique dun foetus
44
Défense, surveillance, sécurité

• Militaire, surveillance de site, ...
• Compression, cryptage, restauration,
• segmentation, reconnaissance

Poursuite davion sur image optronique
45
Global system
Light
f(t)
Sensor
f(x,y)
(pre)processing
matrix g(x,y)
Bases of the world Base de relation
Segmentation
list Eei ei (xi,yj)
Classification Recognition
Graph of objects
Interpretation of the scene
(chaînage arrière)
Reseau semantique
Operation
Data structure
46
Hierarchy of processing

• Low level
• preprocessing, high data volume, low algorithmic
  complexity (filtering, FFT, )
• Medium level
• feature extraction (segmentation)
• High level
• image understanding, low data volume, high
  algorithmic complexity

Image processing
Artificial intelligence (bases, rules)
47
Hierarchy of processing (2)

• Generalised images
• Segmented images
• Geometric representation
• Information models

M flops
Complexity
LL
HL
LL
HL
Complexity depends on what the operator wants !
48
Example
49

• Image Processing
• Image ? Image
• Computer Vision
• Image ? Description
• Computer Graphics
• Description ? Image

50
Tools
No algorithm is done here Softwares Matlab,
Khoros, , C, Fortran...
Computersoftware
result
Data acquisition
51
A step in programmation and computers
PIXEL Picture element is associated to a
value if precision increases ---gt cost of the
file increases (short, float,
double) Calculate 1024x1024 1 byte gt
(0-255) 2 bytes gt (0-65535) 4
bytes more precision colored images 3
images (RGB)
Question Is precision necessary ? medical
applications, ...
52
Cost

• Storage floppy disk, hard disk
  expensive
• Transmission duration of transmission is to
  the of megabytes (web, email)

53
Mathematical tools

• Basic notations and signals
• Transforms (2D)
• Fourier, DCT, Z transform, Walsh Hadamar,
  Wavelets
• Convolution

54
Questions of a teacher

• What is Vision?
• Why is it Important?
• Why is it Difficult?
• Hierarchy of Processing
• Applications

55
Bibliography

• Digital Image Processing, R.C. Gonzalez R.E.
  Woods, Addison Wesley
• Vector quantization, R.M. Gray, IEEE ASSP
  Magazine, april 1984
• Image coding using vector quantization a
  review , N.M. Nasrabadi R.A. King, IEEE Trans.
  On Com., vol36 n8, august 1988.
• Image data compression, A.K. Jain, Proc. Of
  IEEE, vol69 n3, march 1981
• Travaux, thèse et notes personnelles

56
End
Thank you for your attention