UNI

T5
I
NTRODUCTI
ONTOCOMPUTERVI
SION

Computervi
sion i
s a br
anch int
he Domai
n ofAIthatenabl
es
comput
erstoanalyzemeaningf
uli
nfor
mat
ionfr
om i
mages,vi
deos,
andot
hervi
suali
nputs.
.

APPLI
CATI
ONSOFCOMPUTERVI
SION
Thi
sdecadeandtheupcomi
ngonecanwitnessasi
gni
fi
cantl
eap
i
ntechnol
ogythathasputcomput
erv i
siononthepri
ori
tyli
st.
SomecommonusesofComputerVi
sionare:

Faci
alr
ecogni
ti
on
Foreg:Facesecuri
tylock-
indev i
cesandt
raf
fi
ccamer
asar
e
someexampl
esusingfaci
alrecogni
ti
on.
1
Faci
alf
il
ter
s
Moder nday ssoci
almediaappsl i
keSnapchatandI nst
agr
am
usesuchki ndsoftechnol
ogythatext
ractf
aciall
andmarksand
processthem usi
ngAIt ogett
hebestresul
t.

Googl
elens
Tosear chdat
a, GoogleusesComput erv
isi
onforcapt
uri
ngand
analy
singdiff
erentfeatur
esoftheinputimagetothedatabase
ofimagesandt hengivesusthesearch.

Aut
omot
ive
Themachi neryini ndustri
esisnow usi ng Computervi
sion.
Aut
omat edcarsareequi ppedwi t
hsensorsandsoft
warewhi ch
can det
ectt he 360 degr ees ofmov ements det
ermine the
l
ocati
on,detectobjectsandest abli
shthedepthordimensi
ons
oft
hev i
rtualwor
ld.
Foreg:Compani
esl
i
keTesl
aar
enowi
nter
est
edi
ndev
elopi
ngsel
f-
dri
vi
ngcars.

Medi
cal
Imagi
ng
For the l ast decades,computer vi
sion medi
cali magi
ng
appl
ication hasbeen at r
ustwor
thyhelpforphysi
ciansand
doctors.Itcreatesandanal
ysesimagesandhel
psdoctorswi
th
thei
rinterpretat
ion.

2
COMPUTERVI
SIONTASKS

TheApplicat
ionoft hecomputerisper
formedbycert
aint
asks
onthedat aorinputprov
idedbytheusersoitcanpr
ocessand
anal
ysethesituati
onandpr edi
ctt
heoutcome.
Basi
csofI
mages
Thewor
d“pi
xel
”meansapi
ctur
eel
ement
.

Pi
xel
s
Pixel
sarethef
undamentalel
ementofaphot
ogr
aph.
Theyarethesmal
lestuni
tofi
nfor
mationt
hatmakeupapi
ctur
e.

Resoluti
on
Thenumberofpi xel
scoveredinanimageissometi
mescal
led
theresolution
Ter mf orar eacover
edbythepixel
sinconventi
onal
l
yknownas
resol
ution.
Foreg: 1080x720pi xel
sisaresol
uti
ongivingnumber
sof
pixel
sinwi dthandheightoft
hatpict
ure.
 Ame gapixelisamil
li
onpixel
s.

Pi
xel
val
ue
Pi
xelval
uerepr
esentthebright
nessofthepi
xel
.
Ther
angeofapi xel
valuein0-255(
2^8-
1)
wher
e0i st
akenasBl ackornocolourand255i
stakenaswhi
te



































 Gr
ayscal
eImages
 Gr
ayscal
eimagesar
eimageswhi
chhav
ear
angeofshades
ofgraywi thoutapparentcolour.
T hel
ightestshadeiswhi tetotal
presenceofcolouror255
anddar kestcolouri
sbl ackat0.
Inter
medi ateshadesofgr ayhaveequal br
ight
nesslevel
sof
thethreepr i
marycoloursRGB.
T hecomput ersstor
et heimagesweseei nthefor
m ofthese
number s.

RGBcol
our
s
 Al
lthecol
our
edi
magesar
emadeupoft
hreepr
imar
ycol
our
s
Red,GreenandBlue.
 Alltheother col
ourar
e f
ormed by usi ng
theseprimar
y col
oursat di ffer
entpropor
tions.
 Co mputerst
oresRGBImagesinthreedif
fer
entchannels
call
edtheRchannel,Gchannel
andt heBchannel

5
I
mageFeatur
es
 Af
eatureisadescr
ipt
ionofani
mage.

Feat
uresar
ethespeci
fi
cst
ruct
uresi
nthei
magesuchaspoi
nts,
edgesorobj
ect
s.

OpenCVorOpenSour ceComputerVisi
onLibr
aryi sthattool
t
hathel
psacomputert
oextr
actt
hesefeatur
esfr
om t heimages.
I
tiscapabl
eofpr
ocessingi
magesandv i
deostoidenti
fyobj
ect
s,
f
aces,
orev
enhandwr
it
ing.