PERATlON

MODUE S:MoPPHOLO &t CA

Bop set
.Broad uopholpoperotHons hat
tnoceees"eck
hapes e'propertes Sheck
Used for Segmentat¡n kk featune
etralthon

Dpeaatron pixels to boundasle

(i) PoOlation : Adds
ohjech image
rocien Aemee pixels from beurdase
of bject [n_an

ltnctuing Uemert The ne

added oremoe dize 4
Shage of stuctun eleent USedused
shape mast
movphotogmcal operothuns
Can he Ony &hope lk aize d Qach
hes an
 Ivage gnmesh
DILATION ( A B 6

Piys epana
4s o'a binany
GIPands shapes imge
.Au in holes,
gdodlesplzele
Adds an ewtra odtee nha
h bounday making-otyf sghty
ngin
29make
Pixel 1L

engn SE colnidu Lata stuctunng

ele

b4 SE
sE Coingag
olp

f any of válues mateh with

Pe hen the pie
with oill he
eplaced
 vaues hye
with
ma
be
CROSLON>
e(AOstctnng tement
Reoe isolaled noisy plaels
Smoothes obyect boundda
Remores suter layer of p'xels te
ohyect become slighy Siolle
Obect

ogin
4

8eves
Res foes
to euinate noise
poselble oouAmage

lesves to
and holes due to peppe
gel hd of'ghmal nolc
ole
Comoinaton Opening
USed to lean opening cnsing
&emented
betove using'iag % git
HiT MISS TRANSFOR
Detectk a binamy image
a 'stuctumna eleueut taat
Coutain ubinashon
both
gray pixels
eleunE,
Structuny eleneut
Miss: I Strtunng
cowpletely i birealapsoitth. the imiage
contaning bposte pixels valwei tto

conteuniñg
oVertas
Sae Plsel vauea
the ina

atHit: ateaut one pirel vale

utth othe pixel a e se

element b appled to
Gtnychi
Dbtain
ng and
then the ee
Butput iege
 thin 2 Bceletorse
ashape in blnany inage

form !4.

an tteafive pocess where

tn eachteraGn 3ee di fferent
stuctursna eleet to iVsed totdertu
pixet to ko emend

BoUNDARY &TRAtNON

BA) A - (AOB)
ryd Pertorm eoston fixt
tra Bubtrat the EEsion inmage
from
Inede
Resuling inage iaton
 ReBDA DATA onpees sILON
Peduiug the amount of data
eguied to represent a grien
quantity 4 Afor matisn
e: relattve
redundaut data
b'

bt:bits neaded to
f btensity Valies

Ro,1, 2,- 1
MN

L inteity vae
Used foepesent D: ho o timee
intensthes ó
MNimg
intensthy
appes
Pr- pobabilty )
Dceunng j2ig
bitreg to Kd
hp-each'pise
no o b)f Vsed to epat
Rach value Cintensit)
Codingedundony
bodes eae
happan
aleignd to intenstha
cohen
ipithout taking cdvhtag b probabi
intenslhes
ity
Code a
info (8et eents N.
to repvesent eah
&ymbols in code
repeenY
Q-intenity af antmage
harng more bi dn heeed
Yeehese intensittea
o
osless

ReDUNDANy
Related to tnterpixel correlatons
othin an
Value sf awyauny griven pixel can be
redi cted : Aom dlue o t nejghlbouE
the aya intormation caied b
hot Indiridua þiyel þiyel isin yey 'smat
mueh of that pixel edundat
To educe thi
2-D pisel
must be tronstormed aay
to a mnot
ettiuent forrati
between odyacet þtxel dfeene
Can b
Locsles
 Tronsformatons that sredue PR
Cre lalled mOpngs
Caled Deesiblé
ew
mgnal image ele
tel trom transformed ne
etontulte
t)
Ryehoneual ReDUN DA Y /0UANTIZH TION
eninaton f PSYV dataresults t
irexressikle operaton - hence It reAult

eye e boun doent eapond to

au' visual informatan bith eanue
ntenty
&ome hfo s neqleeted dumingroessins
by brain Eliminakonthic ikB
osnt sest affeet intevetaton o
region are iotexpreted
assmp feature 4U broun qpups
correlates' such, Apuptng
mton fo i possible as
ao
 fidelity unteia!

Renongl of irvelevaut date causee

heeded Twoygu
(n obyecte tidaty
fidellty critesia: aoss
eepresed as Mathenatcglfuntion
of e input d output of compreseion
Procers.

S(gnal to noice recio l XL Cxy)

M1 N

Cxi4)-
{(4)
ilp img
-compressiph.
Peso follouled bdeconpresion
(2)2ubj ective fidelity entela i Deeopretd
imoses ae ultimately ieed by humans
lo metu ing innare qualty by
Subjectie
mee pate
Can be one ba Scak
Abeelute rating
de by ide,
lmg uldeo ,audio, Ige
CNe! fsta Vced
sto
Cs, nt but WRoducea
see
e Reduce
t Loupised
galfoina
1t
etereq
to
beCaut hle
to tre d es
he
can e
d d Qre
cOng ae thot that aceAde
btute Rore
thace Keepe eienates
thoun smal
COmpreSRIN osho
loss Ndaa
fomathon
ssy Pregeve
COMPRek&LO IMAGE