3. Images and Graphics - RWTH Aachen

27
3. Images and Graphics Multimedia Kommunikation SS‘99 1 / 27 Contents 3.1 Digital Image Representation 3.2 Image Formats Captured Image Format Stored Image Format Graphics Format Object Representation Bresenham Line 3.3 Image Synthesis 3.4 Graphics System Graphics Hardware Raster Display – Dithering 3.5 Image Analysis Target Detection Digital Image Data Structures Image Transformation 3.6 Image Transmission

Transcript of 3. Images and Graphics - RWTH Aachen

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

1 / 2

7C

onte

nts

3.1

Dig

ital

Im

age

Rep

rese

ntat

ion

3.2

Imag

e Fo

rmat

s–

Cap

ture

d Im

age

For

mat

–S

tore

d Im

age

For

mat

–G

raph

ics

For

mat

–O

bjec

t Rep

rese

ntat

ion

Bre

senh

am L

ine

3.3

Imag

e Sy

nthe

sis

3.4

Gra

phic

s Sy

stem

–G

raph

ics

Har

dwar

e–

Ras

ter

Dis

play

–D

ither

ing

3.5

Imag

e A

naly

sis

–T

arge

t Det

ectio

n–

Dig

ital I

mag

e D

ata

Str

uctu

res

–Im

age

Tra

nsfo

rmat

ion

3.6

Imag

e T

rans

mis

sion

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

2 / 2

7

An

digi

tal i

mag

e is

a s

patia

l rep

rese

ntat

ion

of a

n ob

ject

(2D

, 3D

sce

ne o

r an

othe

r im

age

- re

al o

r vi

rtua

l).

Dig

ital I

mag

e {}

Let

I,J

,KZ

be

a fi

nite

inte

rval

. L

et G

N w

ith G

the

grey

sca

le le

vel (

inte

nsity

val

ue

of a

n pi

ctur

e el

emen

t -

) of t

he im

age.

(1) A

2D

-Im

age

is a

func

tion:

(2) A

3D

-Im

age

is a

func

tion:

(3) I

f th

e fu

nctio

n is

a b

inar

y/b

it im

age,

oth

erw

ise

a pi

xel i

mag

e.

0⊆

⊂∞

×→

××

=

<

pixe

l

fI

JG

fI

JK

G

G

: :

,01

3.1

Dig

ital I

mag

e R

epre

sent

atio

n

The

Res

olut

ion

depe

nds

on th

e gr

anul

arity

of

the

grid

with

pix

els

equa

lly s

pace

d.H

igh

reso

lutio

n im

ages

are

oft

en v

ery

larg

e.

Exa

mpl

e:T

o di

spla

y a

525-

line

tele

visi

on p

ictu

re (

NT

SC)

with

out n

otic

eabl

e de

grad

atio

n w

ith a

Vid

eo G

raph

ics

Arr

ay (

VG

A)

vide

o co

ntro

ller

640x

480

pixe

ls a

nd 2

56 d

iscr

ete

grey

leve

ls g

ives

an

arra

y of

307

.200

8-b

it nu

mbe

rs a

nd a

tota

l of

2.45

7.60

0 bi

ts.

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

3 / 2

73.

2 Im

age

Form

ats

/ Cap

ture

d Im

age

Form

at

Cap

ture

d Im

age

For

mat

is s

peci

fied

by:

spat

ial r

esol

utio

n (p

ixel

s x

pixe

ls)

and

co

lor

enco

ding

(bi

ts p

er p

ixel

)

Exa

mpl

es:

Vid

eoPi

xTM

/ Su

nVid

eoT

M c

ard:

spat

ial r

esol

utio

n:

320

x 24

0 pi

xels

colo

r en

codi

ng:

1-bi

t (bi

nary

imag

e), 8

-bit

(col

or o

r gr

eysc

ale)

24

-bit

(col

or-R

GB

)

Sun

Vid

eoT

M c

ard

(cap

ture

and

com

pres

sion

car

d w

ith 3

0* fp

s (f

ram

es p

er s

econ

d)):

Cel

lB30

fps

MPE

G1

IP f

ram

es17

fps

JPE

G30

fps

Cap

ture

YU

V30

fps

MPE

G1

I fr

ames

30 f

psC

aptu

re R

GB

-8/2

430

/ 12

fps

* A

t lea

st 2

5 fr

ames

per

sec

ond

are

nece

ssar

y if

con

tinu

ous

mot

ion

is to

be

pres

ente

d.

How

eve

r, it

is p

ossi

ble

to “

che

at”

a li

ttle

and

to

wor

k w

ith s

mal

ler

fram

e ra

tes.

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

4 / 2

73.

2 Im

age

Form

ats

/ Sto

red

Imag

e Fo

rmat

A S

tore

d im

age

form

at is

a 2

-dim

ensi

onal

arr

ay o

f va

lues

rep

rese

ntin

g th

eim

age

in a

bit

map

or

pixm

ap, r

espe

ctiv

ely.

The

dat

a of

a b

itmap

is a

bin

ary

digi

t, da

ta in

a p

ixm

ap m

ay b

e a

colle

ctio

n of

:

–3

num

bers

rep

rese

ntin

g th

e in

tens

ities

of

red,

gre

en, a

nd b

lue

com

pone

nts

of th

e co

lor

–3

num

bers

rep

rese

ntin

g th

e in

dice

s to

tabl

es o

f re

d, g

reen

and

blu

e in

tens

ities

sing

le n

umbe

rs a

s in

dex

to a

tabl

e of

col

or tr

iple

s

–si

ngl

e nu

mbe

rs a

s in

dex

to a

ny o

ther

dat

a st

ruct

ures

that

rep

rese

nts

a co

lor

/ -sy

stem

–4

or 5

spe

ctra

l sam

ples fo

r ea

ch c

olor

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

5 / 2

73.

2 Im

age

Form

ats

/ Sto

red

Imag

e Fo

rmat

Eac

h pi

xel m

ay h

ave

addi

tiona

l inf

orm

atio

n i.e

. the

nor

mal

vec

tor

to th

e su

rfac

e.

Furt

her

prop

erti

es o

f im

ages

: wid

th, h

eigh

t, de

pth,

siz

e, v

ersi

on, e

tc.

[In

man

y ca

ses

is d

ata

com

pres

sion

rec

omm

ende

d or

man

dato

ry]

Form

ats:

–R

IFF

(R

esou

rce

Inte

rcha

nge

File

For

mat

) in

clud

ing

form

ats

for

bitm

aps,

vec

tor

draw

ings

, ani

mat

ion,

aud

io a

nd v

ideo

(ex

tens

ion

is B

RIM

)

–G

IF (

Gra

phic

al In

terc

hang

e F

orm

at),

X11

Bitm

ap, S

un R

aste

rfile

s, P

ostS

crip

t,

IRIS

, JP

EG

, TIF

F (

Tag

ged

Imag

e F

ile F

orm

at)

etc.

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

6 / 2

73.

2 Im

age

Form

ats

/ Gra

phic

s Fo

rmat

Gra

phic

s im

age

form

ats

are

spec

ifie

d th

roug

h:

–gr

aphi

cs p

rim

itiv

es: l

ines

, rec

tang

les,

circ

les,

elli

pses

, tex

t str

ings

(2D

), p

olyh

edro

n (3

D)

–at

trib

utes

: lin

e st

yle,

line

wid

th, c

olor

affe

ct.

Gra

phic

s pr

imiti

ves

and

thei

r at

trib

utes

rep

rese

nt a

hig

her

leve

l of a

n im

age

repr

esen

tatio

n. T

hegr

aphi

cs p

acka

ge d

eter

min

es w

hich

prim

itive

s ar

e su

ppor

ted.

Adv

anta

ges:

+R

educ

tion

of th

e gr

aphi

cal i

mag

e da

ta+

Eas

ier

man

ipul

atio

n of

gra

phic

al im

ages

.

Dis

adva

ntag

e:–

Ad

ditio

na

l con

vers

ion

ste

p f

rom

gra

phic

al p

rim

itive

s a

nd

att

ribu

tes

to it

s p

ixe

l re

pre

sen

tatio

n

Form

ats:

·S

RG

P (

Sim

ple

Ra

ste

r G

raph

ics

Pac

kage

), o

ne w

ay c

onve

rsio

n to

bit-

/pix

ma

p

·P

HIC

S (

Pro

gram

mer

’s H

iera

rchi

cal I

nter

activ

e G

raph

ics

Sys

tem

s) a

nd

·G

KS

(G

rap

hic

al K

ern

el S

yste

m)

on

ly im

age

re

pre

sent

atio

n is

in p

ixm

ap

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

7 / 2

7

3.2

Imag

e Fo

rmat

s / O

bjec

t Rep

rese

ntat

ion

Bre

senh

am L

ine

Con

stru

ct a

line

fro

m le

ft to

rig

ht w

ith

a m

ax. g

radi

ent 0

≤ m

≤ 1

, (≤

45°)

,al

l oth

er li

nes

are

achi

eved

thro

ugh

refl

ectio

ns.

Nex

t pix

el is

the

one

in th

e ea

st.

1

d i =

0

••

211

≤+id

Nex

t pix

el is

the

one

in th

e no

rth-

east

.

•••

121 1

1

+=

> +

+

ii

i

yy

d

210

≤<

id

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

8 / 2

7

To

dete

rmin

e th

e ne

xt p

ixel

,

) with

let

the

cont

rol q

uant

ity in

eac

h st

ep,

calc

ulat

ed w

ith:

i

f

if

with

and

if

+1

if

(,

,,

xy

xx

d

dd

m

dm

d dm

yy

xx

y xd

yy

d

yd

ii+

ii

i

i i

i ii

ii

ii

++

++

+ +

=+

=+ +

−≤ >

=

− −=

==

≤ >

11

1

1

1 2 1 2

21

21

01

11 2

11 2

1

10

∆ ∆

>≤

=

>≤∆

−+

+=

∆=

∆∆−

=∆

−=

+++

+0

if

1

0

if

and

0

if

0

if

2

22

=M ,

2)0(

:fo

llow

s

as

of in

stea

d

by

n ca

lcul

atio

Eas

ier

111

1

210

ii

ii

iii

i

ii

ii

D+

y

Dy

yDD

xM

D

MD

D

yx

mx

xD

dD

3.2

Imag

e Fo

rmat

s / O

bjec

t Rep

rese

ntat

ion

Bre

senh

am L

ine

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

9 / 2

7

1

••

ii+

1i+

2i+

3

d id i

+1

d i+

2

Yi

Yi+

1

Yi+

2

Yi+

2

>

−+

=−

+=

+=

+=

+=

+++

+++

21

sinc

e

13

1

2

223

121

i

ii

i

ii

i

ii

d

md

md

d

md

md

d

md

d

≤=

>+

==

++

+

++

++

21

sinc

e

21

sinc

e 1

32

3

21

21

ii

i

ii

i

ii

dy

y

dy

y

yy

Let

Yi =

true

val

ue o

f cu

rve,

then

Yi+

1 =

yi +

di+

1

Zy i

; Y

i ∈

3.2

Imag

e F

orm

ats

/ Obj

ect R

epre

sent

atio

n B

rese

nham

Lin

e

R

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

10 /

27

On

the

corr

ectn

ess

of th

e fo

rmul

ae f

or d

i and

yi:

Yi+

1 =

Yi +

m;

Yi+

1 =

yi +

di+

1

⇒ y

i + d

i+1

= y

i-1

+ d

i + m

⇒ d

i+1

= d

i + m

+ (

y i-1

- y i)

d i + m

if y

i = y

i-1

d i + m

- 1

if y

i = y

i-1+

1=

21:

crite

rion

21

: cr

iteri

on

>←

≤←

ii dd

d i + m

(and

yi =

yi-

1)

d i + m

- 1

(and

yi =

yi-

1 +

1)

21

if

21

if

>≤

ii

dd

=

Cha

nges

of

y i:

y i-1 y i-1+

1y i =

21

if

21

if

>≤

ii

dd

y i+1

=y i y i+

121

if

21

if

11

>≤

++ ii

dd

3.2

Imag

e Fo

rmat

s / O

bjec

t Rep

rese

ntat

ion

Bre

senh

am L

ine

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

11 /

27

Ex

amp

le:

M=

2y

=0

.4,

D0

mx

y

x

=→

==

=−

=−

02

10

2

1

.,

.,

∆∆

∆∆

xi

y Di i

12

34

56

78

910

1112

1314

1516

17

00

01

11

11

22

22

23

33

3

-1-0

,60

-0,2

00,

20-1

,40

-1,0

0-0

,60

-0,2

00,

20-1

,40

-1,0

0-0

,60

-0,2

00,

20-1

,40

-1,0

0-0

,60

x i

1

2

3

4

5

6

7

8

9

10

1

1

12

13

14

1

5

16

17

1234

y i

3.2

Imag

e Fo

rmat

s / O

bjec

t Rep

rese

ntat

ion

Bre

senh

am L

ine

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

12 /

273.

3 Im

age

Synt

hesi

s

Imag

e sy

nthe

sis

(gen

erat

ion)

is a

n in

tegr

al p

art o

f al

l com

pute

r gr

aphi

cal u

ser

inte

rfac

es a

nd in

disp

ensa

ble

for

visu

alis

ing

2D, 3

D a

nd h

ighe

r di

men

sion

al o

bjec

ts.

E.g

.:

Gra

phic

al U

ser

Inte

rfac

e (G

UI)

: des

ktop

win

dow

s sy

stem

with

icon

s an

d m

enu

item

s

– O

ffic

e A

utom

atio

n an

d E

lect

roni

c Pu

blis

hing

: des

ktop

pub

lishi

ng, H

yper

med

ia s

yste

ms

– S

imul

atio

n an

d A

nim

atio

n fo

r S

cien

tific

Vis

ualis

atio

n an

d E

nter

tain

men

t

Pict

ures

can

be

dyna

mic

ally

var

ied

by a

djus

ting

the

anim

atio

n sp

eed,

por

tion

of th

eto

tal s

cene

in v

iew

, the

am

ount

of

deta

ils s

how

n et

c.

–M

otio

n D

ynam

ics:

Obj

ects

are

mov

ed a

nd e

nabl

ed w

ith r

espe

ct to

a s

tatio

nary

or

also

dyna

mic

obs

erve

r, e

.g. f

light

sim

ulat

or.

–U

pdat

e D

ynam

ics:

Obj

ects

bei

ng v

iew

ed a

re c

hang

ed in

sha

pe, c

olor

, or

othe

r pr

oper

ties,

e.g.

def

orm

atio

n of

an

in-f

light

aer

opla

ne s

truc

ture

.

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

13 /

273.

4 G

raph

ics

Syst

ems

/ Com

pone

nts

of in

tera

ctiv

e gr

aphi

cs s

yste

ms

App

licat

ion

mod

el–

repr

esen

ts d

ata

or o

bjec

ts to

be

pict

ured

(st

ored

in a

n ap

plic

atio

n da

taba

se)

–st

ores

gra

phic

s im

age

form

ats

and

conn

ectiv

ity r

elat

ions

hips

of t

he c

ompo

nent

s

–sh

ould

be

appl

icat

ion

spec

ific

and

inde

pend

ent o

f any

par

ticul

ar d

ispl

ay s

yste

m

–co

nver

ts im

age

data

base

rep

rese

ntat

ions

to th

e gr

aphi

cs s

yste

m fo

rmat

App

licat

ion

prog

ram

–ha

ndle

s us

er in

puts

by

send

ing

com

man

ds to

the

grap

hics

sys

tem

des

crib

ing

wha

t to

disp

lay

and h

ow th

is o

bjec

ts s

houl

d ap

pear

Gra

phic

s sy

stem

–in

term

edia

ry c

ompo

nent

bet

wee

n ap

plic

atio

n pr

ogra

ms

and

the

disp

lay

–e

ffect

s an

out

put t

rans

form

atio

n fr

om o

bjec

ts in

the

appl

icat

ion

mod

el

–e

ffect

s an

inpu

t tra

nsfo

rmat

ion

from

use

r ac

tions

to a

pplic

atio

n

–co

nsis

ts o

f out

put s

ubro

utin

es c

olle

cted

in a

gra

phic

s pa

ckag

e to

dis

play

imag

es

Gra

phic

s ha

rdw

are

–re

ceiv

es in

put f

rom

inte

ract

ion

devi

ces

and

outp

uts

imag

es to

dis

play

dev

ice

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

14 /

27

3.4

Gra

phic

s Sy

stem

s / G

raph

ics

Har

dwar

e

Inpu

t:

mou

se, k

eybo

ard,

dat

a ta

blet

, tou

ch-s

ensi

tive

pane

l on

the

scre

en (

2D in

put)

trac

k-ba

lls, s

pace

-bal

ls, d

ata

glov

e et

c. (

3D a

nd h

ighe

r-di

men

sion

al in

put)

Out

put:

ras

ter

disp

lay

Com

pute

r H

ost I

nter

face

Dis

play

Con

trol

ler

(DC

)

Vid

eo C

ontr

olle

r11

0110

0001

0100

1010

001

0100

0101

0100

0000

1000

000

0000

0001

0000

1000

011

1010

1101

1100

0001

0000

100

1000

0001

0010

0000

000

0000

0010

0010

0000

1010

101

0111

1110

0001

0100

000

Ref

resh

Buf

fer

Dis

play

Com

man

ds

Inte

ract

ion

Dat

a

Mou

se

Key

boar

d

Scan

ned

Imag

e

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

15 /

27

3.4

Gra

phic

s Sy

stem

s / R

aste

r D

ispl

ay

Sta

rt r

aste

r sc

anra

ster

line

s

vert

ical

ret

race

pixe

l

hori

zont

al r

etra

ce

To

avoi

d fl

icke

ring

of

imag

es a

60

Hz

or h

ighe

r re

fres

h ra

te is

rec

omm

ende

d.

At e

ach

pixe

l, th

e be

am’s

inte

nsity

is s

et to

ref

lect

the

pixe

l’s in

tens

ity.

In c

olo

r sc

reen

s th

ree

beam

s ar

e co

ntro

lled

(red

, gre

en, b

lue)

.

Mon

ochr

ome

disp

lays

use

ac

hrom

atic

ligh

t tha

t is

dete

rmin

ed b

y th

e qu

ality

of t

he li

ght

(inte

nsity

and

lum

inan

ce).

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

16 /

273.

4 G

raph

ics

Syst

ems

/ Dith

erin

g

Dis

pers

ed-d

ot o

rder

ed d

ithe

ring

(m

onoc

hrom

e di

ther

ing)

is u

sed

e.g.

to e

xten

d th

e nu

mbe

r of

avai

labl

e co

lors

of

disp

lays

(at

the

expe

nse

of r

esol

utio

n).

Exa

mpl

e:C

onsi

der

3 bi

ts (

red,

gre

en, b

lue)

per

pix

el a

nd a

2 x

2 p

atte

rn a

rea:

Eac

h pa

ttern

can

dis

play

5 in

tens

ities

for

eac

h co

lor

resu

lting

in 5

x 5

x 5

= 1

25 c

olor

s.

Hal

fton

ing

(clu

ster

ed-d

ot o

rder

ed d

ithe

ring

) us

es th

e hu

man

eye

’s c

apab

ility

for

spat

ial

inte

grat

ion

to p

rodu

ce d

iffer

ent i

nten

sity

leve

ls, e

.g. u

sed

by la

ser

prin

ters

that

are

not

abl

e to

disp

lay

indi

vidu

al d

ots.

Exa

mpl

e: 2

x 2

dith

er p

atte

rns

(5 in

tens

ity le

vels

) re

pres

ente

d in

dith

er m

atric

es

How

to “

prod

uce”

mor

e th

an tw

o in

tens

ity le

vels

if o

nly

two-

leve

l int

ensi

ties

may

be

repr

esen

ted?

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

17 /

27

3.5

Imag

e A

naly

sis

Pixe

l inf

orm

atio

n is

ver

y of

ten

insu

ffic

ient

! T

here

is n

o in

form

atio

n ab

out:

–sh

ape

of a

n ob

ject

–or

ient

atio

n of

an

obje

ct–

thre

e-di

men

sion

al in

terp

reta

tion

of a

n ob

ject

Í Im

age

Ana

lysi

s (

Imag

e P

roce

ssin

g) is

req

uire

d.S

ome

appl

icat

ions

or

spec

ial t

opic

s ar

e:–

Imag

e en

hanc

emen

t–

Pat

tern

rec

ogni

tion

–S

cene

ana

lysi

s–

Com

pute

r vi

sion

Exa

mpl

e:

Tra

ffic

scen

es ta

ken

by a

cam

era

inst

alle

d in

a c

ar

Pro

blem

s:Is

ther

e a

traf

fic s

ign

visi

ble?

W

hich

tra

ffic

sign

?

Is a

mov

ing

car

in fr

ont o

f our

car

?

W

hich

type

of c

ar?

W

hich

rel

ativ

e sp

eed

to o

ur s

peed

?

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

18 /

27

Imag

e an

alys

is is

con

cern

ed w

ith te

chni

ques

for

ext

ract

ing

desc

ript

ions

abo

ut im

ages

:–

com

pu

tatio

n o

f p

erc

eiv

ed

bri

ght

ness

and

col

or

–p

art

ial o

r co

mp

lete

re

cove

ry o

f 3D

da

ta in

a s

cene

–lo

catio

n of

dis

cont

inu

itie

s co

rre

spon

din

g t

o o

bje

cts

in a

sce

ne

–ch

ara

cte

risa

tion

of

the

pro

pe

rtie

s of

un

iform

re

gio

ns

in a

ima

ge

Imag

e E

nhan

cem

ent

–im

pro

ves

ima

ge

qu

alit

y b

y e

limin

atin

g n

ois

e (

ext

ran

eo

us

or

mis

sin

g p

ixe

ls)

or

by e

nha

ncin

gco

ntr

ast

, i.e

. X

-ra

y im

ag

es,

com

pu

teri

zed

axi

al t

om

ogra

ph

y (C

AT

)

Sce

ne A

naly

sis

and

Com

pute

r V

isio

n

–d

ea

ls w

ith r

eco

gn

isin

g a

nd

reco

nstr

uctin

g 3

D m

od

els

of a

sce

ne

fro

m s

eve

ral 2

D im

ag

es,

i.e

.in

dust

rial r

obot

se

nsin

g (r

ela

tive

size

s, s

hape

s, p

ositi

ons,

col

ors)

Pat

tern

Det

ectio

n an

d R

ecog

nitio

n

–d

eals

with

det

ectin

g a

nd c

lari

fyin

g st

anda

rd p

att

erns

and

find

s di

stor

tions

from

thes

e pa

tte

rns,

i.e.

Opt

ical

Cha

ract

er R

ecog

nitio

n (O

CR

)

–st

atic

cha

ract

er r

ecog

niti

on (

OC

R),

dyna

mic

rec

ogni

tion

(ha

ndw

ritin

g)

3.5

Imag

e A

naly

sis

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

19 /

27

0000

0000

0000

0000

0000

0000

1111

1111

1111

1111

1110

0000

0000

0000

0000

0000

0000

0000

000

Tur

ning

rad

ar d

etec

tor:

an o

bjec

t

0000

0001

0000

1100

0000

0000

1111

1100

1101

1111

1001

0000

0000

0100

0001

1100

0000

0000

000

Impe

rfec

t rad

ar: s

ome

0-1

mis

take

s:

Elim

inat

ion

of m

ista

kes

by d

igit

al f

ilter

ing:

Sim

ples

t ver

sion

: sy

mm

etri

cal (

2m-1

, m)

filt

er

0000

0100

0010

0011

1011

0110

1111

0001

0

0000

0000

0000

0011

1111

1111

1111

0000

0Fi

lteri

ng b

y (5

,3)-

filte

rE

xam

ple:

=′

′′

′′

→∑+

−+

=

else

0

a

if

1

with

...

......

...

1-m

h

1i

21

21

mh

in

ni

ni

ma

aa

aa

aa

aa

3.5

Imag

e A

naly

sis

/ Tar

get D

etec

tion

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

20 /

27

Sour

ce o

fIm

age

Tar

get

Imag

e

Sym

boli

c R

epre

sent

atio

n

Con

ditio

ning

Lab

ellin

g

Gro

upin

g

Ext

ract

ing

Mat

chin

g

Form

attin

g

Step

s

digi

tal i

mag

eda

ta s

truc

ture

logi

cal d

ata

stru

ctur

e

unim

prov

ed

impr

oved3.

5 Im

age

Ana

lysi

s / I

mag

e R

ecog

nitio

n St

eps

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

21 /

273.

5 Im

age

Ana

lysi

s / D

igita

l Im

age

Dat

a St

ruct

ures

Form

attin

g

–ca

ptur

ing

of a

n im

age

and

tran

sfor

min

g to

a d

igita

l rep

rese

ntat

ion

Con

ditio

ning

–ba

sed

on a

mod

el th

at a

ssum

es th

at th

e ob

serv

ed im

age

is c

ompo

sed

of a

n in

form

ativ

epa

ttern

mod

ified

by

unin

tere

stin

g va

riatio

ns–

est

imat

es in

form

ativ

e pa

ttern

on

the

basi

s of

the

obse

rved

imag

e–

supp

ress

es n

oise

and

per

form

bac

kgro

und

norm

alis

atio

n by

sup

pres

sing

uni

nter

estin

gsy

stem

atic

or

patte

rned

var

iatio

ns–

app

lied

unifo

rmly

and

con

text

-inde

pend

ent

Lab

ellin

g–

base

d on

a m

odel

that

ass

umes

that

the

info

rmat

ive

patte

rn h

as s

truc

ture

as

a sp

atia

lar

rang

emen

t of e

vent

s–

dete

rmin

es in

wha

t kin

d of

spa

tial e

vent

s ea

ch p

ixel

par

ticip

ates

–la

belli

ng o

pera

tion:

•ed

ge d

etec

tion

, cor

ner

dete

ctio

n,•

thre

shol

ding

(e

.g. f

ilte

rs s

ign

ifica

nt e

dg

es

an

d la

be

ls t

he

m),

•id

enti

fica

tion

of p

ixel

s th

at

part

icip

ate

in v

ario

us s

hape

pri

miti

ves

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

22 /

27

3.5

Imag

e A

naly

sis

/ Im

age

Tra

nsfo

rmat

ion

Dig

ital I

mag

e D

ata

Stru

ctur

es a

re a

chie

ved

thro

ugh

imag

e tr

ansf

orm

atio

ns:

Let

I,J

,G a

nd

an

imag

e.

:

is

an

imag

e tr

ansf

orm

atio

n.

⊆×

→×

×

Νf

IJ

G

GG

IJ

IJ

:

τ

Exa

mpl

es:

1. L

ocat

ion

depe

nden

t wea

keni

ng o

f In

tens

ity (

e.g.

eff

ect w

ith le

ns):

2. T

hres

hold

-Tra

nsfo

rmat

ion

(e.g

. filt

erin

g of

gre

ysca

les/

colo

rs):

τ (f

)x

yg

xy

fx

yin

tens

ityf

unct

ion

xy

(,

)(

,)

(,

)(

,)

==

+

Let

a

nd

the

thr

esho

ld,

an

imag

e an

d

(he

re

) is

def

ined

as:

fo

r

for

so

tha

t

is

the

new

His

togr

am o

f d

efin

ed a

s:

wit

h n

umbe

r of

pix

els

wit

h gr

eyva

lue

Gr

rs

G

fI

JG

tG

GG

:=

,

ti

is

si

r

hh

hG

hi

fi

i

tf

f

ff

:{

,...

,},

: :’

’{

}

(): :

()

()

.

()

=∈

∈×

→→ =

≤≤

<≤

→=

=−

0

01

0 1

0

1

Ν

Ν

1

rs

t(i)

it(

i)

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

23 /

273.

5 Im

age

Ana

lysi

s / I

mag

e T

rans

form

atio

n

Exa

mpl

e:

3. H

isto

gram

Spr

eadi

ng (

e.g.

to in

crea

se th

e co

ntra

st o

f an

imag

e)

[]r

it

h

ri

b

bi

a

ai

r

ai

ab

ri

t

Gt

rb

ia

f

GJ

If

rr

G

ft

,0)

( H

isto

gram

,

new

th

eis

that

so

for

for

0fo

r

)(0

:)(

used

.

are

of v

alue

sgr

eysc

ale

w

hole

the

follo

ws,

as

defi

ned

W

ith

.0

of sc

ale

in

the

sgr

ey v

alue

only

uses

imag

e

the

Ass

ume

imag

e.an

:

and

},

,...,

0{:

Let

)(

∈≤

≤<

<≤

−=

<≤

≤<

→×

Ν∈

=

Ana

logo

us to

spr

eadi

ng th

egr

eysc

ale

can

beco

mpr

esse

d an

d us

ed o

nly

in p

artia

l int

erva

ls.

A c

ombi

natio

n of

bot

h is

e.g.

use

ful t

o m

anip

ulat

efi

lm m

ater

ials

.

r

ra

b

Ran

ge o

f ne

wgr

eysc

ale

valu

es

Ran

ge o

f ol

d va

lues

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

24 /

27

3.5

Imag

e A

naly

sis

/ Gro

upin

g an

d E

xtra

ctin

g

Gro

upin

g:–

iden

tifie

s ev

ents

by

colle

ctin

g or

iden

tifyi

ng m

axim

al c

onne

cted

set

s of

pix

els

part

icip

atin

g in

the

sam

e ki

nd o

f eve

nt (

neur

al n

etw

orks

!)

–de

term

ines

new

set

s of

ent

ities

–ch

ange

s th

e lo

gica

l dat

a st

ruct

ure

–e

ntiti

es o

f int

eres

t afte

r gr

oupi

ng a

re s

ets

of p

ixel

s

–e

.g. li

ne-f

itti

ng is

a g

roup

ing

oper

atio

n, w

here

edg

es a

re g

roup

ed in

to li

nes

Ext

ract

ing:

–co

mpu

tes

for

each

gro

up o

f pix

els

a lis

t of p

rope

rtie

s:

–ce

ntro

id, a

rea,

orie

ntat

ion,

spa

tial m

omen

ts, g

rey

tone

mom

ents

, spa

tial-g

rey

tone

mom

ents

, circ

umsc

ribin

g ci

rcle

, ins

crib

ing

circ

le, n

umbe

r of

hol

es in

a r

egio

n, a

vera

ge

curv

atur

e in

an

arc,

etc

.

–m

easu

res

topo

logi

cal o

r sp

atia

l rel

atio

nshi

ps b

etw

een

two

or m

ore

grou

ping

s, i.

e. c

larif

ies

whe

ther

two

grou

ping

s to

uch,

are

spa

tially

clo

se o

r la

yere

d

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

25 /

27

3.5

Imag

e A

naly

sis

/ Mat

chin

g

Mat

chin

g:–

dete

rmin

es th

e in

terp

reta

tion

of s

ome

rela

ted

set o

f im

age

even

ts r

ecog

nise

d pr

evio

usly

with

the

extr

actin

g st

ep

–a

ssoc

iate

s ev

ents

with

som

e gi

ven

3D o

bjec

ts o

r 2D

sha

pes

–te

mpl

ate

mat

chin

g is

a c

lass

ical

exa

mpl

e of

a w

ide

varie

ty o

f mat

chin

g op

erat

ions

,

com

pare

s ex

amin

ed p

atte

rn w

ith k

now

n an

d st

ored

mod

els

and

choo

ses

the

best

mat

ch

Con

clus

ion:

–C

ondi

tioni

ng, l

abel

ling,

gro

upin

g, e

xtra

ctin

g an

d m

atch

ing

cons

titut

e a

cano

nica

l

deco

mpo

sitio

n of

the

imag

e re

cogn

ition

pro

blem

.

–E

ach

step

pre

pare

s an

d tr

ansf

orm

s th

e da

ta to

faci

litat

e th

e ne

xt s

tep.

–O

n an

y le

vel t

he tr

ansf

orm

atio

n is

an

unit

proc

ess

and

data

are

pre

pare

d fo

r th

e un

it

tran

sfor

mat

ion

to th

e ne

xt h

ighe

r le

vel.

–D

epen

ding

on

the

appl

icat

ion,

the

sequ

ence

of s

teps

has

mor

e th

an o

ne le

vel o

f

reco

gniti

on a

nd d

escr

iptio

n pr

oces

s.

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

26 /

273.

5 Im

age

Ana

lysi

s / D

etec

tion

of li

cens

e pl

ates

Form

attin

g

Con

ditio

ning

Lab

ellin

g

Edg

es

Ext

ract

ing

Gro

upin

g

a gr

oup

Mat

chin

g

Aso

me

puri

fica

tion

has

been

don

e

stru

ctur

es a

s co

rner

s, e

dges

...

are

dete

cted

prop

ertie

s as

siz

e,or

ient

atio

n ...

are

dete

cted M

eani

ng d

etec

ted:

It is

lette

r A

max

imal

gro

up o

f ev

ents

3. I

mag

es a

nd G

raph

ics

Multimedia Kommunikation SS‘99

27 /

273.

6 Im

age

Tra

nsm

issi

on

Exa

min

atio

n of

the

netw

ork

for

imag

e tr

ansm

issi

on:

–ne

twor

k m

ust a

ccom

mod

ate

burs

ty d

ata

tran

spor

t due

to la

rge

size

of i

mag

es a

nd d

iffer

ent

com

pres

sion

res

ults

–re

liabl

e tr

ansp

ort o

ver

the

netw

ork

is r

equi

red

–tim

e de

pend

ence

is n

ot a

dom

inan

t cha

ract

eris

tic o

f the

imag

e (in

con

tras

t to

audi

o an

d

vide

o)

Imag

e re

pres

enta

tion

form

at u

sed

for

tran

smis

sion

:

–ra

w im

age:

Imag

e ge

nera

ted

thro

ugh

the

vide

o di

gitiz

er a

nd tr

ansf

orm

ed in

its

digi

tal

form

at (s

ize

= s

pati

al_r

esol

utio

n x

pixe

l_qu

anti

zati

on),

e.g

. 640

x 4

80 p

ixel

s an

d

quan

tizat

ion

of 8

bits

req

uire

s tr

ansm

issi

on o

f 307

.200

byt

es

–co

mpr

esse

d im

age:

Dow

nsiz

ed im

age

with

met

hods

suc

h as

JP

EG

or

MP

EG

.

–sy

mbo

lic

imag

e: Im

age

repr

esen

ted

thro

ugh

sym

bolic

dat

a as

imag

e pr

imiti

ves,

attr

ibut

es

and

othe

r co

ntro

l inf

orm

atio

n (e

.g. c

ompu

ter

grap

hics

, im

age

= [r

ed b

ackg

roun

d, y

ello

w

circ

le o

f 5 c

m r

adiu

s w

ith c

ente

r po

int(

x,y)

, ...)

.