DigitalImageProcessing(DIP):
Introduc6onandFundamentals
A
Prof.PascalMatsakis
DIP:Introduc-onandFundamentals I.OriginsofDIP
Prof.PascalMatsakis
Digitalpictureproducedin1921
OriginsofDIP I.1.NewspaperIndustry(1920s)
3
Prof.PascalMatsakis
Firstpictureofthemoon
byaUSspacecra@in1964
OriginsofDIP I.2.SpaceProgram(1960s)
4
Prof.PascalMatsakis
DIP:Introduc-onandFundamentals II.InterestinDIP
Prof.PascalMatsakis
Applica6onsinmanyfields:
medicine,geography,physics,astronomy,defense…
Examples:
characterrecogniEon,fingerprintrecogniEon,
targetrecogniEon,weatherpredicEon,crop
assessment,detecEonofbonefractures,
detecEonofbraintumors,studyof
high-energyplasmas…
InterestinDIP II.1.Overview
6
Prof.PascalMatsakis
II.2a.ImagingintheElectromagne6cSpectrum
7InterestinDIP
λν =c, E=hνEnergyEFrequencyνWavelengthλ
RADIO WAVES
MICROWAVES
INFRARED ULTRAVIOLET
VISI
BLE
GAMMA RAYS SOFT X-RAYS
HARD X-RAYS
400nm500nm600nm700nm
Prof.PascalMatsakis
Bonescan
(tolocalizeinfecEonsandtumors)
II.2b.ImagingintheElectromagne6cSpectrum
8InterestinDIP
Gamma-Ray
Prof.PascalMatsakis
II.2c.ImagingintheElectromagne6cSpectrum
Imageofacircuitboard
(todetectflaws)
9InterestinDIP
X-Ray
Prof.PascalMatsakis
Fluorescencemicroscopeimagesofcorn
(tofindifcornisinfectedbysmut)
II.2d.ImagingintheElectromagne6cSpectrum
10InterestinDIP
Ultraviolet
Prof.PascalMatsakis
II.2e.ImagingintheElectromagne6cSpectrum
Imageofpapercurrency
(totrackandidenEfybills)
11InterestinDIP
VisualSpectrum
Prof.PascalMatsakis
SatelliteimageofNorthAmerica
(globalinventoryofhumanseRlements)
II.2f.ImagingintheElectromagne6cSpectrum
12InterestinDIP
Infrared
Prof.PascalMatsakis
MRIimageofahumanknee
(diagnosisofsport-relatedinjuries)
II.2g.ImagingintheElectromagne6cSpectrum
13InterestinDIP
Radio-Band
Prof.PascalMatsakis
II.2h.ImagingintheElectromagne6cSpectrum
14InterestinDIP
480-710nm
Panchroma6c
Prof.PascalMatsakis
440-490nm
520-570
II.2i.ImagingintheElectromagne6cSpectrum
15InterestinDIP
630-740
Colour
Prof.PascalMatsakis
450-515nm
525-605
630-690
II.2j.ImagingintheElectromagne6cSpectrum
16InterestinDIP
775-900
1550-1750
10400-12500
2090-2350
Prof.PascalMatsakis
450-515nm
525-605
630-690
II.2k.ImagingintheElectromagne6cSpectrum
17InterestinDIP
775-900
1550-1750
10400-12500
2090-2350
FalseColour
Prof.PascalMatsakis
II.3a.OtherImagingModali6es
Imageofanunbornbaby
(todeterminethehealthofhisdevelopment)
18InterestinDIP
Ultrasound
Prof.PascalMatsakis
II.3b.OtherImagingModali6es
19InterestinDIP
Computer-generatedimageofahumanskull
(forcriminalforensics)
Synthe6c
Prof.PascalMatsakis
DIP:Introduc-onandFundamentals III.MainStepsinDIP
Prof.PascalMatsakis
IMAG
E PR
EPRO
CESS
ING
L O W L E V E L
H I G H L E V E L
outputsareimages
5 REPRESENTATION AND DESCRIPTION
outputsareimageaRributes
III.1a.Overview
21MainStepsinDIP
IMAGE RESTORATION
2
IMAGE ENHANCEMENT
3
KNOWLEDGE BASE
1 IMAGE ACQUISITION
4 SEGMENTATION
6 OBJECT RECOGNITION
Prof.PascalMatsakis
LOWLEVEL:
acquisiEonandimprovement
ofpictorialinformaEonfor
humaninterpretaEon
andanalysis
HIGHLEVEL:
autonomousmachinepercepEon
III.1b.Overview
22MainStepsinDIP
Prof.PascalMatsakis
DIPcharacterizedbyspecificsoluEons.
III.1c.Overview
23MainStepsinDIP
Techniquesthatworkwellinonearea
canbetotallyinadequateinanother.
ActualsoluEonofaspecificproblem
generallysEllrequiressignificant
researchanddevelopment.
Prof.PascalMatsakis
III.2a.Examples
24MainStepsinDIP
Whichstep?
Prof.PascalMatsakis
III.2b.Examples
25MainStepsinDIP
Whichstep?
Prof.PascalMatsakis
III.2c.Examples
26MainStepsinDIP
Whichstep?
Prof.PascalMatsakis
Whichstep?
III.2d.Examples
27MainStepsinDIP
Prof.PascalMatsakis
DIP:Introduc-onandFundamentals IV.ImageDefini6onandRepresenta6on
Prof.PascalMatsakis
ImageDefini-onandRepresenta-onIV.1a.AnalogImage
29
fa | 2 → R [0;+∞[ funcEon
funcEon
fa | 3 → R R
image
fa | 2 → R R3 funcEon
funcEon
fa | 2 → R C
fa|→ isatypicalcase;R2 [0;+∞[
fa(x,y)istheintensityoff
aat(x,y).
Prof.PascalMatsakis
x
y
ImageDefini-onandRepresenta-onIV.1b.AnalogImage
30
Considerfa|→R2 [0;+∞[
z=fa(x,y)
x
y
z
M
N
0
L
DomainofdefiniEon[0,M]x[0,N]andrange[0,L]
y
x
M
N0
Prof.PascalMatsakis
ImageDefini-onandRepresenta-onIV.1c.AnalogImage
31
fa
ZeroPadding
Prof.PascalMatsakis
ImageDefini-onandRepresenta-onIV.1d.AnalogImage
32
fa
CircularIndexing
Prof.PascalMatsakis
ImageDefini-onandRepresenta-onIV.1e.AnalogImage
33
ReflectedIndexing
fa
Prof.PascalMatsakis
ImageDefini-onandRepresenta-onIV.2a.DigitalImage
34
((x,y),f(x,y))
f | Z 2 → 0..+∞
funcEon
funcEon
f | Z 3 → Z image
funcEon
f | Z 2 → Z3
f|→ isatypicalcaseZ2 0..+∞ :grayscaleimage
loca6on graylevel pixel
Prof.PascalMatsakis
f(0,0)f(0,1)…f(0,N-1)
f(1,0)
. .. .. .
f(M-1,0)…f(M-1,N-1)
ImageDefini-onandRepresenta-onIV.2b.DigitalImage
35
0
1
2
12...y...N-1
0≤f(x,y)≤L-1
M-1
x
...
...
pixellocaEon
Considerf|→Z2 0..+∞
Assumerangeincludedin0..L-1:
,i.e.,l -bitgrayscaleimage typically,L=2l
AssumedomainofdefiniEonis0..M-1x0..N-1:
Prof.PascalMatsakis
ImageDefini-onandRepresenta-onIV.2c.DigitalImage
36
y
xpixel
f(0,0)f(0,1)…f(0,N-1)
f(1,0)
. .. .. .
f(M-1,0)…f(M-1,N-1)
Considerf|→Z2 0..+∞
Assumerangeincludedin0..L-1:
,i.e.,l -bitgrayscaleimage typically,L=2l
AssumedomainofdefiniEonis0..M-1x0..N-1:
Prof.PascalMatsakis
ImageDefini-onandRepresenta-onIV.3a.Digi6za6on
37
analog
digital
digi6za6on
sampling
quan6za6on
0..+∞
fa | 2 → R [0;+∞[
f | →
Z 2
→ [0;+∞[ R
2 → 0..+∞
Z 2
Prof.PascalMatsakis
1024 512 256
128 64 32
sampling
ImageDefini-onandRepresenta-onIV.3b.Digi6za6on
38
Checkerboardeffect:
Prof.PascalMatsakis
32 16 8
4 2
quanEzaEon
ImageDefini-onandRepresenta-onIV.3c.Digi6za6on
39
False
contouring:
Prof.PascalMatsakis