0% found this document useful (0 votes)
112 views160 pagesDigital Image Processing
Uploaded by
Neha ChogaleCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
112 views160 pagesDigital Image Processing
Uploaded by
Neha ChogaleCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 160
co! Sigs ans Sot. To anfersand image funamentls and taasfoms|
image enone! nus pail and eine domain To tay i
ra cep gl inne proceing 9a. Ahmet
revo value he techies for nage enbarcereat and images tt
ass iage processing techniques for si Introduction to Image
Processing
‘eying 0 aay them and dissing it wit bers
the be, Refer Fi 121
> tena 20 signa ana Sysome-—~ A et snow defi wt we mean by an age. TH
word eund is mensional, wie mages tied
a Heace an image cn Be
In this case, to locate the do.
Al family piace, photographs on iy cane
‘© 2imensonal. is stent acer, ete
single example
‘Bcamplof«Idimenonl an «2-dimensonl signal
1 ws ee why images are
en
onside te irage shown in Fig.133.
Fig. 134
shown, went
The sesromepetic secon cag be expel
leewomgncis radiation. One cannot exist witout te | tn f wavelength nd quency. The weet (ad
fer A ow of leaie comet sinaye proses
Imagoetsm, and magetisn is wed w produce ec.
leeuomagostc aan i propagated outwards om is
source at 2 velosity of 300000000 meters per eco
x10" whee,
ee a emit to «spectrum etenting trom 380m 0
Most fhe mages hat we encounter incur ay ody
|) ire take fom camers which ar sensitive to thi ngs
ofthe electromagnetic pectum (380-760 nn), We mst
ot fore hough, that tre re cameras which are cpale
(te fen (0a reo ye expen
ant aay
Here ci te pest ght 3310! ee
us fequecy of tcintons can be elated sing
uation (14.
a ndesand the chutes of the me,
Fr ecient design of algorithns win
r
L
ree stone the ge alc te Bnd se)
15 _Units of ntensi
Weknow tat foran object to be seen, ome amount |
eas foo
Tee nit of Tumioos imensiy () candela (66)
(SI ei) wich by cefaion, is (V50)* of & square
center ofthe sue of 3 Bick body radio te |
hole empera of 2045 K.
[hte A Back boy enor sone a bso a
‘adeton der Joon and hae 90 reeceng
power Te race tom a rasind Bick body
sexe clas ck by con ai aways
‘he uni is called camila because inially was
ened a + sancard cane baring pce amount of
a per bur There is snter inp ni apr fran
‘tel tt we encounter. Tits cll a.
aise. clea is the white potion of they
Inside he chro i the etna which is composed of
mo pes of photoreceptors Thee phokreepirs re
Spelled io covert ght rays Ino receptor petals
‘These photoreceptors ee called ros and cones ound due
1o be shape, Rods ae long and slender wl cones ae
sorter and thicker, Nerves consecting the sens lene he
‘eh dough the ope nerve bundle. Light entering the
come i foci onthe retina surface by alent changes
pe under muscular eal opefoem proper focusing of
Cones provide colour vison an Ag vial city
(shy of visio) a bight eh Cones eigen higher
level of amine, tir apne clled Phonic
‘ison. nt interme sarge of lumii, bth rote
sax come reacties prvi Meope Viin,
a moon ight we camo seco bene only he
ros ae functioning, Te distin of rode an cone in
the ena estown Fg 162
be .
Wen + ight sims sees 4 ra of» cove, &
owoctemietanston cca, prodaing ane ip
‘The timer in Which te tere inpbes propee
‘trough te vial sytem a eb ally ude’ ty
"be mea sty an ance et wed or dcusion oo
e astomy of be humane ee. Ar ted eat
Geo en oo way exhsive aed nero det,
can refer modal jour, But at sade of nage
rocesing hi informivon woud be cent.
116.1. Image Formation in Eye
ee The bei atl ook hi as i,
the op. At he cere of tis was ale, te ity
‘unaan is inetd tom 1101+ A. The eben
row sed bere wheter He OF SE can dete
incre
he cere noo dette change, the nen
is fbr incesed by another increment of AL Ty,
procedure is continued il He observer detects gy
trea, Ths is Known the Jt Nodceahe Dita,
ov)
fa 34mm
The beight of erin nage 34
he same image is obser at sta 10m,
laoet AS woud Wec ate
ay
Weve raio = 4h
‘Te human ee isa complex eaten andthe images
ta we perceive ae ase equly comple. We al am)
explain evisu phesoneat
What the hamsn ee sees we i genera inesy
image Ines, Baghoces nd Contat are dee
irae phenomena. Intenso ht care deeds
te ta aout of it ee by Hence ints
ly propery and caste mes ihe
she hand i eyo visual concept an ene seca 3
seo o ih meas. Cones maybe defied as he
Atfereac in pecivd bigness.
A) Coates Sensi: The respons ft buman eye
{changes inthe nea of iui is kxown
be nomtna
Conse te single exernen:
‘0 time our discrimination quality also reduces whet
here isto ch ight
intensity incident ate aan ee.
‘Te care in the Big 164 repre the rnge of
Intsies hat be human isl se can adap.
‘Wien ilunisatn i ow, i te
Pays dominant cole wile fr i
level of minaion sod ten ngs st geting
isles aed,
(©) Acaty and Contour (Mach Bands): Mich de
tamed afer the Aun physct Et Mach
mina.
[As canbe sen fom he fg, the son fom
seatpic vison to photopic vision is gradual and st
ceraia eves of uintions bath fem pay le,
To cat a lng sary sort, the dynamic ange of he
stan i left. This eke Mach a effect
‘Teac and he peeved nest cars a shown
lng. 168 @)ant 0, |
($) Integration : Anode important property of gy,
ha it inegaes the Sene as a Whole,
rope hi aiats proesmig, When
tea pice, we dono 0k teach point, bt
c et
ssaviole
Leta an example oe ing sg
Wa ome matn by nae poeeng?
|
ee
L
(0 Atal itensty =
a
Dhnctin’
Aecodrng oa reste at an Blogs Unser,
Cae eniacy nctaten ta ht
Sronee drninaion ie poor fw eso
inmaon Espa
ts cater, petininay concep of i eee iy
procesing ae oie, Difewnce srtormache perception. Ean o
ne dimensional sd wos signs is x
inspite of te speing mistakes. 2.8 wat coe ne te to
(Peeve inensity You cn ead 0 Topics sucha eevosgstic pcm and ive eaeenaeeed
Fie ba ‘eau of te integration proven ofthe ee.
uw a dtcused with exanples. Elements of te taman | 0-2 What do Mach bonds inp?
ve) isa stem ae presenied, Bas astm ofthe hana | 0.10
Wan oben ching an jc 2 ig are
incase teed 40 overtook atthe sight hand
ee 7 9€ explined with a few iourwions. Percept ince of 2m Fi et Pa
trates each as righes adaptasion an lope ™
irguoh betioan nage proctting and
aos
2.12 Enis to tem brane asgatn,
‘base requie io undentnd nage processing
‘Chapter Bas.
aaa
(thematic
function)
cy.
Esch of he smal icles have the same intensiy, but
fecase the surounding grey level ofeach of the
From the Table 16.1, 21 _ Introduction
7 Basle slements of an image processing
system
Digi image rocesing i sll ean of
Seo comp. Te a ee ge Plone + Pcoric devies comit of | 4 nln of CCD pie a cal otto be
reg ym sown low, Quantum detectors sxmlondacer jc, Thay ae sit at ars | pect eter than 213 Tozer
ee ‘compote dcr lize nag nents town | cups watage etl pepo! be omne
ne eg aaa ntinractosducient 5 Fhomstm, Fame anes poe t mae "OW foe. One hig st te ay COD
Image Sg. steed phat bang med nspenonceeeee ee, OAM ial ati popes weeny cee solo ie (ec oie}. re
©) nage Presi eae ear ae aes ane oe on cee oe ee acceee eee trie a eroding, th ae aray CD
(©) Dp. Woenver tis cca, te properties ofthat maul pp ccf ponte doves ager tas bee cing device y vig
sterd in some meatutle way, Plnc/Enstin came a score hc sone mechs avy.
(6) Trannies ‘it clan erwen th wavelength of he nciar TH RSMO we in solid-state iging snot
Perper bwsed penipaly on chapecowped devs, cmeny | Tay eetigr ir ed in ft bed seamen
West cus ch oma et amecanel tow a Chard Conped Devies CCDs. Hesce tt | samen tat joe coos aca in your boy we
a (21) imegng esos cae OCD sean exter ale Aline ray CCD st ve ayhigfm fow
Te slid se aay (CCD) cn be araegd in to =
Weer, 6 = ny cared ya photon a
Pct cont, 626% 10% 5 (Lin ary CCD (2) Aes amay CCD.
© = Speedof tig, 3% 10% mice (dL Arrays: The ine are eres he inp
er so ha te ice he
4A = Waren oft incident raion pare ae
Pig 21 Or collision, the photon ruasfers all or none of! ay of phone,
‘uantum ofeerz othe elec, mapa ets
(0 mage Acq: ngs eso ite it tp
U2 SY image procesing system. The general aim of | ___TH Equation (2.1.1) is very important a it tls LS ss vee
pata i fe maxima wvelegth o which the goat ‘ i
wil ipod is deermined by the coergy vest! t
ene by te mal elec,
citreus
Ime etn witty |g MH mae pn
we dfect cameras for dite applications I ve manana
(9 Mande Pwemic
Pe yes wth ne aye a
(Aran Array : The pot
rc only one ie, To get a teodimensnol HPA
ees mechanically move te aay ve ee
iver
{UL
Gh
tft
Fens
Lp
‘Ara amays of music rays const of
wo inensional any of poe
ashe wives eal word sees wit
toy mecbaial seaming. Trot much more
information can be daluced from a single rea ine
iaace tan would be posible with line args. Area
“aryscan te see i al te dpi cameras that we se
fox video imaging. Te ara aay ae more verti
than the ie aay, a ee BS epg
iis Area ays ne higher om co abd compli,
presen come in Ee EES. 2565 6
“90 x 380, 640 x 480, 780% 575. CCD arrays ge
icy pskaged as TV camer A sence,
te ee a ery igh pet (10.00 sn
ries tm lf appeaions im Which fei
etn eu
(a) mae Storage + Al video signals a een
alg form ie, elec signals convey Huminay
td exiour with cotavnsly variable Voge. Th
ances ae lnerfced (© 2 computer where te
poccsng algrthas ae wenea, This is dove by.
frame grabber car. Usually fame grabber cri
Prine ciel Board (PCB) Fite othe os comps
(256 to 4056 gray level). Hence a frame grabber cet
isan interface between the camera and the compe:
bate)
canon [0] 2
Fig 216 Simplified dagram ofa frame grabber card
&
aol athe Form bytes (write no
frame gbber memory unde he compet sally
by DMA tnasien, The cinents of the menexy are
cotinuosly read ou a eo re (0 fees), ped
hugh the DIA comer and spayed onthe mone.
‘The ouput bas color map of «cole Look Up
‘Tuble LUT to pei roo cooing. The ble coasts
Wit he advent ofthe PCT bs, he hot compute
memory could now be Wed a he Rane Baler emery
‘This is ecu he PCT bus faites ft common
Hence te irage could nom dsnly be sored on the
compete RAM Duet this, tb fae baer memory bas
ecome or ores let Oly the vey igh nd fame
or | Biber cade sl Save sol frame enbter memery
‘he grey level at hat pou. Suppose
the image (mai) sof ine Ax B and spone we egie
© bis to represent each elemear of te mat. Meniey
space required fo sor this image i proximaly equal ©
AXB Cb
Suppose we have D such images hen tol umber of
212)
Fie 200)
meled onthe.
Iags being rw dimensional fntons, 0cpy aot
of pace an ence is ave ove a compu With
Sele ard is and ab a good RAM. Proesed images
cull be soe oe mage Noes or CDs. Archival ta
se od on maga pes
(Image Procening + Systems engine from
ricco 1 gener puro Ire compuien
sae wed im image pressing. Debeaed image
rossi systems comes te Bet ompues are
‘very poplar owas rcesng of iil images
lnvole process tat ae wally expesed in ee
—_ a
oe —
63S ‘f=
a)
mee —
‘The image seacce forthe Aner sce it a concise representation a possible for transmission over
coc
Mos ssi i rds tlevison ae el in analog form ad analog cing ethos ae sed to make itz
nsession hanes a cocoa pie incltig cable tea ato, slies, and opal fibres.
amen: abe neg raison cin bere ae many elements which exit x noolinear behaviour, If xis ea
sey pn tee
yea?
1 Gans oft dese
pe ee ey ee ee te nity peas we aie) Heme mae
ZEEE EE te Gotu coms cesar new ao conpesaing orn
Wetec tof igs fares vale at ths age to ne rnc agra ofthese topes
22 _ Image Types
1 was sated eater har images are mensional
fictions Retr Fig 221
ages canbe classed as follows
(0), Monoetome ges (Binary ines)
(2) Grey scale images,
@) Colour 249 images.
© Hatranet aes
Pig 2a
images. I such images, we have nly back a white
pines and no ober sbaes of ey. Refer Fig. 222.
Feit
(2) Grey Seale Image : Hee cach pie is ay sored hi fe sss Bey 8 e949 pas
‘hoe pt aac
fom
oar ig cb ey comer
Lpseig sap seni)
The has oes he te
Mol) as it conmind ir convenicg im 8 ge en, Cine
‘Acute ta otc sinires | gay 20500 eh or te ga
(Hat Toning: vis tw ey see image
eiey oaks tener ane incon image
snes mere grey ees. But er a oie in
nd Mat of te pst tt we He Gj er,
in Grotas cau spy en x ain ar al Sere devs, hey tae ely
“ 2 bk cries ean only pce two eel
che (lick a ne bck oud). fa mt of te
a igh aoe ing vel devices.
MATLAB cae orceering ok image
ie mela matic of ee ew eating
Peaay
ey lee equi wef be mash whet pa
teste so camp
You tive all ad senspapers a some pot of ie
opel). The images ook he ey level imagen
my.228
ean be see hit os at ok he wey
Me a smh tak oe. Hace Tes su te
‘oe whe cfg he mae,
eli arcu tsi i ack fr ey evel,
Fig 226 ‘Te fog to inpeent haloes age oa ey
Pig 224
ispiven ow,
rete ft bck represents white, last ck
‘ven the images ht you seein mos of the books
| SEBGME hice ae psec wg Wack elu 4
‘rie biclerund ln ite of his we d get an iio of
Ye se of he hone mates bed
‘tuner of grey eves he oil irae
we encour grey evel 0, iy ie | mgs can dy te efor O88 pig ©
an ater Howe BPS ies Be don
conti pier.
0 te Poppe Exe O00) He a
fe eames dt deep a ae fort ye
et opin ai
{0 images compre nase Wich cap,
ery ite sce Hi ted che Dist Coe
TawtomDCT (xpined io cher of Ig,
efor) JPEG imps ae JOY compre
Airis vio ein soe os of rgd ag
ene te uly JPEG rage ct good as By
Atop Free ern etnies, hy
stare vasa,
Tie Fig. 23:1 stows he opie opeiaion
a eal pt comeing a
eens he py sof he haem Inge
23 _Image Fle Formats
tp ldo este ase
ompnr ingles ee A fom
ch fs ow neato edi te Ganon
Mose 9 poms a
een rap) Fp Son
Tepe en) OE ngaero
eB 203
Wha 2a
fe ROB aes intra h vaes fne
ary ok for al pat spe then te vale
"he expiry clu owed yh ves the
imay car. Ths he mage aa ee or,
RRRR GOGO. BBB Bn
‘The comprsion pe ine hae nies wheter
‘te image din is compose ing lin ach Ra
Jeng eeoding «Apu fom tee, re ae & few mare
fonmas which we ome soe while dealing whim.
POM (Prue Grey Map) -POM sed
esa py evel images.
'PBM (Pore Bit Map -PBM is wed erst iney
PPM (Ponale Pel Me) - PPM ie wed represen RGB
These fomats are dingsitied ty 2 charter
MP ie i ery gd ut ey xy
of wenory space. PICT is 4 genta pape ows
es Sit cour and 24.81 color imagen. TH!
fs th erin whic pine valet
Stomtre | ime ype _ | trae pe
me P| Biowycran | asc
72 | oreyseie Pav] _ascu
The age omits ow
= a PB ROB, ASCTL
(Beaks 2) tage ts piel | caveman ‘het he ie ofthe image 8 x8 ad eam have 255 grey
Peer levels ach ales eps by Bt
+e se inrmaion aout te Kin fail ocean Ee
Inthe row ampli) we th tet pe level 5% aig an i
“encprimciinst
era
Now, comes te obvious question As we krow, moe
fe sal ad eB Deter hems, What tn
shoul be the eal valbes of sampling and quanta,
‘This aver wil vary fom image to image. Given below
‘ampling and the quantization vals. Aste
te qumization increase, the numberof Hs
rth image increases tremendously
‘The clanty increases, but sorage space requind
We ence need o et a uade-off between he
Mig. 332,
For simplicity, we consider a square mati of st
Comparing the images, we se tht “ls contouring”
Fi=(255jmaxime(F); % sorting ‘ake place at we rece the number of grey levels,
MATLAB code for edcing spi resolution
‘46 Down sampling
pa ‘%P%To se the llc of reducing he nner of wpe 5%
imahowinB(F)) leo
-
RE easement J= 19 Thin ah eae he ae on E Rae eaee
nei fhe second image asa “Checker Boar” PAE fe
pis doco of amples.
~ 34__Isopreterence Curves
We hve see te effect of
ceding pi. We sil dono know
OF N Gamplingrescuton) ad m
by TS. Huang i 1965 steopes, 3.5 Physical Resolution
ly the effects on imsge quay Sar
and
By 00 we know data digi! imag, or image for
shor, is composed of discrete pixels. T \
‘A 640% 480 image woul measure 646 aces by
itches when delayed
jective quay
ere summarised in the form of isoprene
ee ‘Ts 23% 2 ich image at 4 relation of 30 pinls
‘Per inch would ave anal of $4000 pixel)
a
Aitowh tis coavetin mies it mle) ig | pape 0 0
oudepene
Sc ee 249i porn peo
cing oy
—_—
eae
rear 1 Tae on ai pie
nese et a engl ina, a ea
chile stdin
son
Wetsow
ence theese image wl be 60048.
mies
‘commer mea of rae fr itl da is te ba
defies 6 ember of Bs sated per sod.
(0) Now if we tse faster wansmisin ie of 7:
tse Bi, we a BV 8 ag
+2) etl
N= tone 1004010
1 = 1osssrensie
‘Tes ie anid 56 avd.
‘Time taken = 1872510
‘Time taken totus the imate
‘Terfore to represeat this image, we
2as2 20m pita 2.4 expan nonar srtng
8 compute me pry!
when pnt ya tr
tease
‘Weta 4CT image wich we was wit ro
elon in. The CT age sf siz S12 312 a a
256 gyros Dui tnaeision each ble peel:
a and scented by 4 wp Bt How many mt wi
ew tani ine?
(ole pase eps nes an ent 9600 ice)
Sole:
256 grey level required mean bit sini
“£
4.1 _ Introduction
Image nhac one of te fs sp ne
merely improves the subject qual of the images by
working with be exsing data.
Imag enhancement ca be doen to dais.
() The spatisl dooain
2) The feguecy domain,
Image Enhancement in
Spatial Domain
42 _ Spatial Domain Methods _
em pa doa ans wii Be ng
Meo ge nls ee wh
| Slower wre wig Se i em
1a fa 99 ig mie wf Be
st ven (9a ape conti For
tng an ewes fom 0 255 we Oe
tte 285 mpc well Be emi Ya
rpms.
8626 sown in F421
ry”
ew
he igi image ad T is te
(© © ae anew meio image
{8 9-Fer al pido einge t sol ps
‘hinge. The seer equsin remit ae,
‘To sumtin, spl dowain tgs a toe,
which Grey work wit te pel vas to gota von
Image based on Eaton (42),
Spatial domain caancemeat aa be cared ot in two
feet ways
(Point processing
@)Neighboutond processing
Ho)
(256,256)
Leto stat with expaing wha is meant by the
‘pail domain, sass the vain metods of sail
domain enancement sed tan move on to dices the
frequency domuia techague ia chaper of image
etanceneat in Frequency Doi.
Pipa
Note : We always consider the axis a6 shown is
Fig 421, The rani is taken in the downward ved
ction, while the y-aris is taken in the hore
irction. The reason for taking this inverted coor
sytem isto make sue tat the image uses the stnitt
‘mtrx coordinates that We have been using right fom
‘shool. You will remember te fist vale ofthe mai?
‘ways atthe top let corer while the last value of
‘matin isthe ight botom carer.
Fiend
43 Point Processing
point processing, we work with sale pixels Tis
151 operate. It means ta the new vale fa, ) deeds
(on the operator T and the preset 9). Ti sitemet wil
be clearas west giving ome exan
‘Some of te common examples of pint processing re
(1) ig nezaive
an |
oxgeus, native simply means ivering the grey
levels lack
Ooi graylovers 85
Fig 4.2 = 0.54 255and when = 255.8 #0.
Inger,
fe aepr aay
Here Lis te namber of prey eves. (256 in isos)
(MATLAB program fr Ring te gal negative
I MATLAB cet excl naive 5%
se-red (not
e=dinbefan)
£255 ora Bi mage
=
Seu)
crm)
Snot)
ee
nag Exhncemann Ste Don,
(@)Digal negative
Fig433
2) Contrast stretching ¢ Many times We obtain Ing
ctrat images doe 10 por uminations OF doe
wrong sting of the lens aperture. Te Ide behig
contrast stetching ist inrease the contrast of
mages by making the dark portions darker and
Iwi portions brie.
Fig. 634 shows he transformation used 1 achiew
comma steching. In the Fig. 4.3.4, the doted Ine
depending onthe input image and the application, As
was mentioned, image enhancement is & subjective
mn et ing en
43
‘Wore £m and m a the slopes. is ler fon the
‘igre tat and nae es ha one whem ie peter ta
one. The contastsueching tsirmation inreses the
“ami range ofthe mode nage.
MATLAB program forconts srehing
59 Co ting a anise 6
5 Slopes ake es 0.5, 2nd 0.5%
earl,
as
ev immead¢ apr)
ada
Low catenseo,
LE-input( eth wer toatl vale’);
‘UT~ inp Ene the upper thea vale)
mUELN,
Is
(0) Contrast stretched mage
Maas
©) Thresholding : Exeme costa sbeching yisks
‘treshoding. WF we observe the coats suciching
‘Gingram loely we note tht if the fr andthe Tsk
‘slope ae made zr, and he cent lope is icresed,
(Original tmage
‘The forma for sieving threshing ia follows
rsa}
ifr>a) mas
igi gry leva
Fla 4246: Original grey bevel
‘Where Lis the numberof grey level
‘As meatoned eater, image entancement being &
subjective phenomena, te valve of a will vary tom image
to image and from person to person. The abecive 6 0
iesity the region tat he or she i ineresed ia. An
Iimporan hing o ote i hata hesolded image has the
maximum contrast at it as only back and whe grey
ves oan! yore
(a) Sting without background
os
aig=o,
gure (1); 9 — ginal igs,
imshowio)
ot
gare) Mom ry ee acing with agra
=
gu (15% gia ings
imahow (p)
Figure 2), 9 gry level scing without background
43.9 imshow (ii)
This metod is known as Grey eve slicing witht
Wekgun This is because in this process, we ve
‘comple os the background. In some applications,
tony eto enancea band of grey levels but also se!
() Gray eel sticing
—— voreda,
es ree a ene me aed
—
‘ bk td) ‘means 28 or 256 grey levels,
si
SESH as settee eT pacity to
“ clear all ‘example black is represemied as 00000000 and white is. ed pln sein. |
eh a a ae Te
ere a he 8 ages |
ine il ot ilo ings | ssemio wae tt 3
v5
Since 7 te mama py level met iy i | end
a mpmaet ne ey i
ace we iv 38 ps, Corin eae |
toting Wg,
Biny image LSB ise Mie it
pie plane | Eeggah 0
MAT AB ol ect g K
— "Se
4 hee
:
£
jo | Ned
a
Pi 42.11: gh igs, each ry a
| ‘onto seat
Observing the images we come tthe conchsio
|e thr oner tis coin maoriy of te nad)
Gente in while the lower bits comin test
Si te age. Bit pane sing can tence ve vel?
Caee
Pip 431s
[MATLAB program for dysamic range compression
9% Dynan range comprestion 8
D0 Nea te cer he
436)
low al coed ang image peri.
Se Sano cpm De |
sang i Ng
Once ets 9 i aed, we it he ma by one
sep wo
‘te Fh to the tp Now WS einer
theqvncies in
‘imeesonl signa. Given two signs, we can eaily
Singh Yetcen the ower feuecy sgn and the
‘Ther ae 4 et of things that can be achieved by | she foueey signal ReerFg 4,
‘ighouroodpiceaig which nt posible with pit
4 mask oF a window =a
| SN Soccer
Sx Sor a7 x 7..) mask nae
nthe image, malipy each component of
corresponding vale of he image ad hem vp.
Sa,
‘This operation isthe same as convolation, In the
aes Where we consider one piel 188
J cob ato wea 5.004737... neigh
te voltages change i an ndtion ofthe frequeney,
‘The same concepts can be appli to images. i images,
"stead of vues, we have grey levels. If the grey levels
, | change sow
hangs, ar high regueacy areas. Always reser i,
7 sts asa
ove ity com ii
442 Noise
sony oft ie, ey acids
Raise
Expense
OF tee,
chen, We lexi the Gaus asa ad pepe
oi itis caper,
1) Gwsineie—(@)_ Satan pepe
po Gayle!
a Mean average va of
co Sad devon
vein
BERETA.)
oti val it ABE 20), 435
sin noe hs 8 BAKINED VAIS Aan
ar lig of
(uc) Some books ee totes peck noe
‘Toke te same igs the one taken fr the Guan | rom the pa regen we geet he mask that
camp the low past fiwing eration. One
i gt te rm he epi reponse it hata
— ; | |
sre ere conon an th
T b Grey level
Fig 4.410(0)
When ak and pepper creeps ie the imae looks ike
|
re Aas
‘Thee ar mo constant regions in the image
Pig, 44100)
noi ceen i. Che BPE St and paper ise res
$: Gas oe ee qu tame ch aay swing tke ple | 488 se inact hat he mae hs Sn wooo en®
09 9 9 5
Jomonnn 0 0
‘hi ema twa eight il we ec te
line and ten ove downed. Some of te mst
sos we so be.
[pelniniaiaiss)
[esesseee)
neem eny,
pee mony,
ey
may
serene,
|| 02 2m ee,
won nenn,
lonuonwe
benwenee
nooeeeea|
[n20 22989) |
Mow: The estrous be win wa vow mame
|
‘Theat ow inte seven bing
he poy ea changes © 23.33 ar rounding of, fom
Bitetangst0 366 (6 ate runt off ant fly fom
2 hangs Wo 50, Heoce we can sy th the ap ge
ar tecme lure. Chek for yourself wha wu happen
you took a bigger sy, 55. (ou wil ave ake
bigger nage ote youre)
‘Tes kindof eran ies a excl’ when the
Image cosine Gain oie. K ashieve fkesing by
uring te sie. Some of the ct lw atk averaging
sks a shove.
0
0
0
0
0
0
2
”
0
10 0 w
0 0 10
10 10 1 10
wo 10 10
2 0 9 90
» 9
0
0
0
so 50
0|0 9
0 0|9 9
‘The reat of convolving the image wid the 3x3
eraging as i how,
nm © wo w w
op ww w 0 w
oe ow w 0
as ma moms ms m3 ms
HS me Ms O68 Be oss
oe ae © » w
oo 2 0 © 2 »
soe » w so
ajo
MATLAB program fr low pa fering
he images with Guise
5 Lo ps filer wed ona image with Gumi sie
neal
os
tained) sn 0% 00,
doubles:
imma puta) ding cine eon
prerer}
[fs]
»
(Orn! ag ith
(0) Lom psd ise
reas
Genera MATLAB progas rio paseid
sing sy mk soe
15 Lowy erg a image ig verging asym bleyh
sable (91.2)
inn
”
‘ie (Low Pas Peed Inge);
| ecamziams
Mle oP Fier a ing MATLAB,
44.5, Low Pass Median Fitering
‘The merging fer eover te se by Hung it
‘10 Jager sen. Bain th rcs aloo rs
es. Big th averaging mack meres te Nanay.
‘Ther ce times when the img conti Sl and pepre
se. we wo an averaging tert remove the sane it
wil rte ete butt would lo rine edge Hee
‘when we asd to eliminate at nd pepper aie, ve wee
ith nonincr fier known the Median fier They we
ho called orders iter bctise thie pone
sel onthe onering oF ranking of the pixels conned
dh he mask
4 this case, we we x mak smart the eveaging
‘ier except th te musk as 9 valu So ier ke
‘working rectly with te 8sghbous fhe cen pte,
‘Theses to perio median tering te flows:
(1) Assume 43%: empty mas.
@) Place te empty mask athe ef hand comer,
(©) Araage the 9 piel in ascending or descending or
(4) Choos the median fom tee nine values
(Pace thie medina te ceo,
(6), Move the ras ina slr fahion othe aveceing
fier
1m median tering, the grey level ofthe cet pine is
‘replaced bythe media vale ofthe neighbourhood. Always
rit the resultant ina new matrix (image) We stall etplain
‘median fering wit an example, Inte image shown below
et 250 be the salt nd peper nose. I we us an weraging
fier, tbe nose would yread out andthe edges would aso
cod up geting bured. I we only want to remove the noise
Without dsurbing the ees, we use a median fe
0 10 0 0 0
9 0 0% 0 20 0
on 20H 9»
9 0 50 0 50 50 50 50
500 50 50 50 50 50 50
on he th ean cae, we tat by acing he mask
1 10
0 0 © 0 0 20 © 50
00 990 % 5
0 0 5050 0 © 0 50
0 0 9 0 0 0 3 50
0 10
29 90 0 0 0 0
no DDH 0
0 0 00 0 50 50
o 9 09 0 9 0 0 lunonn 00x
ee eanonern
= me au te edges. Median fiers give
See yhee the amoant of mise i
ee ow
0 0 0 0 50 50 50
ad wo f-101), pera ee
ino te ne of be mk i 1% 3 ad te term O indicates the poston from whe
= (10 1 singly ens a te ie
‘sng stars
Ci ooven emmy 4 [EET] onion toss
syoeuen? Me
1 wv vo
Brot “2
fee eRe e aKa | pear ———— ? 7
pea = i
fk fom)
ima (py 5 | Imtonsin
priest mine eum) i
ate) Ibe td
SL
(Original image with) Medan fered image
ine ‘at aod pepper mine
Jed) he ith aan ibe re
paaieeeememne re Pe 4a
45 _Highpass Filterin
Fig. 45.
‘As seen fom the spatial respons, it clear tha the mask coefficients have wo be sah hat they
the centre and negative values a the periphery.
(One ofthe high pass masks is shown ia Fig. 45.1,
‘The important thing to pote i thatthe sum of the coefficients of he high pass mask has 10 be equal wo aero, Tiss
‘because, when we place this mask over the low frequency regions, the result shoul be zr. cornea
eigen rch a
To Semen tee
ee ema isi ati
per tnscaag na tm
een ia
mone te Seo mn ts
uk x wold se OE OBES
2 2701 2 ata i ya
oni isnot ter tng 8
nee. 2 fe 0 ak te Md alas
Conve 7 i ey
een va Wt Wl hen
ema a wl we rege!
wig 8 EE Yb
tee = A WO a
rien 0 al pve vale be yj
Flt wes
ea of fe ri OHS eee
ag nt ir 1 4 IY Fe gt
rae dhe we ere WR 8. We
cocounter this problem in practice. To eliaay 4
rose, weet amastWi AsaingFincn,
ence fr pastel PUOSES WE SE mag
0 0 0 0 0 OOO
oo 0 0 0 0 0 0
0 0 0 0 0 0 oO O
-20 270-2 -20 -2m0 -210 270 -20
am mo || Nt 19 is simply scaling fonction wes
20 210 2m Mm m0 70
°
o
low pass mask in which the 19 term is & pat?
oe oe © © || averaging operation. In this case we could also wt #
0 0 0 0 o © 0 | | U8 017 depending on the image. Some of the oe
eee ae
Seeoneere| (elle) flat
renee :
lafefa] [afals
nwentemupenaoet tne
‘nee Pil aes comet be neu. Tey ays
San wih 260. Here We sed 6 ged of te
Te rel ing the fist mask and pune
(0) High pas tered image
Fig 452"
4.8 _High-Boost Filtering
(4 Original image
MATLAB on arth pu ig
Wis ang TTS
fou
‘As can be sen the high ps ier gives great ess,
But ther is one problem. ges id of the complete
backround, There af tmes, When We eed oeahance te
‘ges bot a retin some ofthe background. To dota, we
se & movi veson of the igh pus fier known as
igh Bos ering
Jn High Boos ering, we pass some of the
‘wcground along wih he high reuency cote.
‘Weknow tat High pas = Orginal —Low pass
‘To pas some of the background, we mulily the
igh Boost = High pass
A> 1 then some ofthe original signal iad back
‘background into the high passed image. This technique it
tno known as unsarp masking. ‘Te mask coetiients for
igh bost tering ae show beow,
Hee X= 9-1
Henceif A= 1,X=8 whichis highpass mask,
ENN ete weld edna ater,
MATLAB program for high bons iting
{1 Hi oe ering
erat
Heh be
(Ore at to
fiero sng ay mask sizes given next
5: High pe hing an iage 5%
He | ara
‘he geerined MATLAB program for th
as) mab tiga 3)
od
tay) <0
bi=0
bow(it)
eH Pe Pier Tage)
47 Zooming
Acoter ingorant spin in image eahacemest
‘woes the sual domain neighbourtond operation wed
image zoming. Zoomlag of mages snot someting that
"ew you. Ifyou have used Mir pint or Phctoshop
editor, you would be aware a there is an option which
allows 0s to z00m an image by 25%, $O%, and 100%
Have you ever imaged at bow at iage that,
‘ook small can soealy look so big The technige is
‘uit simple and ater eading the next couple f page, you
‘woul be ale o doi youre
Zooming cn be cared out wing two feet methods,
(Replication (2) tmepoaion
‘We wil discon both these metho in deta, We shall,
St start with zooming ung replication.
ears
ATA Replication
4 replication, we simpy replicate cach pine and thea 10203040
io own #820 interlacing [
4.7.2. Linear interpolation
In tis method, insead of repiaing each inc,
average ofthe to adjacent pitts log the rows i ken
and placed between the two pitel. The sue operation ia
‘hen performed slong the colames. Tis operation i one
nthe intetaced mage,
$00 a he ep ane cour of oe 4
eae othe a raced age
ee ‘As encond err, rong by meplion ive
we ‘ ‘ial img «phy ok ce cates apy lr ae
fore. This cnt thal de,
‘todo toning kao inl
472 Linear interpolation
ing wheter
1 ts etd, asad of mpeg each pve,
se of he 80 ace sog te er ster
2 np ng,
eta rraapon i en
evap
taslis sacs
crane ntuneiinte
os
toma
11 YW 10 Uh had
Rec the a ie in
&
Zooming by
(Org) Zoom
re 472
44_savetssempies
an te a sie BO
nowt ti tage Hence te
Soin.
© Contrast stretching r,= 5,15 #3 26,8, =2:
aii
toy =[s[ifale
slifel2
ala|sle
‘The conta wweching esfomation it shown in
Pe F483,
‘Rom be ates of and, we ca id te slope of
eves that is image can Bold ie 2? a3
256,
eae he isin prey evel is 0
‘he be maimam prey eels 255,
1a inlar mune, we can find te vale of Busing he
vest 068 A») (2) Bit plane ateing
‘soc te image si, = 2? = 8
We convert the given imags into Waky ad chee
‘epante he ple. Since the given images 3.8, he ley
4 1
i Lf [os [os] (0) Treaoling wie T= 4 - : we
1 1 7
Eee Te tresdig futon i shown blow aba °
Wee mt ot ae ELEY ei a ofol fo} [efits [a] [ifololi 7
‘ ieee An MsB plane —Cenerbitple LSB plane ‘
tap ; (W) Clipping with r, = 2,195 *
Tre ciping wamsornaion is shown in t
Fe Pana), SMT
©
Fig. F485 Continued... |
RBS rete Th be el ig te Gey lng
‘rman wich ive by,
“alsel a]
‘When we rely he LPF on he nag, he cee pia
changes
Witou tkgrine
nae
Witton
Hae K + Origa numberof grey levels
- Moitied suber of grey ees pent nr arn tnenents)
ey His ing mp 1S We ea
iy aye de ay gece
255 Ge imma vee
san
nis, we 270 pa the image before petoming te
th tsnsien, | ferns option Th 66 age
Original - Low pass =
ne mask overs image tartng IF
cordiin 0 Be ving ing,
cen nen co see on
ose fa ae im, H
ten
nis we mpi te Dk i © eH
WAN EEG (mein. 2-)
oeaier
- a ttm, np
Qs: | a
T ™ T 7 =
Far. 42100)
Theta it eos tenn enn
i tin
i | 7
Me Pao) ‘ony Low nd gh Ps Spt soe he owing Fi, P4812 mai Prove ha High pn = Onin - Low pt. Asse
ita Rows a Colm,
SN
tape
Leh ath mmf
coo
methine | meraaiia Newmar
meoeifan re |
| ‘Zero ping F, and F, so tat both are of at
(0444-1) x(0, 4-1) 934
e
°
fax = [0
°
° gages) BO)
ed
ih patty) =
i= fOg BN) =
‘Wesel High pss =
Tee gh pas ea Teco,
Hays =
aut | m7 | -0667
[esi [ ta | 2238
Onin Low pas
x] [x] [an
1066 | 2366
= [33 | 0 | 90/-| sos | 41.22 | 30.66
sofa {a} [ae | soae | 23.667
CS
Fd ner conv pt ime, 3) th eo 9
Lette opt
step
(he ctr methodist rahi tecig),
We zero pad fy and, 9) a ach way both
of bem eof sa, ewosoncsets
san er =r
@
‘Tefal image wou,
for itensiy level sig ise
ey
© ei el gm i,
(pe et 88 lS je
0) Ne |
Usa us eves 2 and $e mae
‘ait Bening ves remain uachaneo 3"
fea
1
ofofolr
‘The LSB and MS planes ae ston below.
ofolsfo
tfolofo
ifs
fof
°
ofthis
LSB plane SB pane
s{s[alo
2{als[e
s}a}ilo
49 Comparison of Contrast Stretching and Thresholding
Sr
Na.
‘Contrast Stretching
‘Threshokiing
Contrast srtching increas tbe dynamic
| ten te tiveshld are made black wile
‘Treesolding ako increase that dynamic
‘ange by setting a tvesbold. Al values ess
Values greater than the theaold are made
qua to white,
The formula of the contras-stetching
‘Uzi sven below
‘The formula for achieving thresholding is as
f 17 Low pater is soning fe Explain
10 lve be scout of enhancement ter athe
tal domain,
|Belkn why modan Mer et stab for guslan
roe, What se timton of mean hor
ivan a kage, what i he ovpet ung a 8x 3
sroragh fer end a medin ther? Catan te
How Js image subracton carted cut ? Gh an
‘ample of image sutvecion in the Yea ot
‘redtine.
(2.26 Exclain the majo erence betwen hgh pas and
‘woh boost rng win an example,
(G27 Dacre arn oat of rpenedy apn
Prepares beittesntccony
7 | 2.90 The gey levis nan mage range om 10 0 50 an
\we wah pie helmage on Soviet has &
{20y lovl rang of Ot 265. Wine tanstormeton
2.31 Fotowing io Is an nage wih 8 gry level ‘This i known as & nomalied istogrum, Toe
stvanag f te sctnd maths i tat te matin aes
tobe plots wil aye he L. Gane ick i coed
1 ey lve! Oa whe ah matin uty king
te hisogrm ofthe nage re! dato latomation
cm be obi Some of he pial hogan ar stows
ve guy of see Be ri
ge Homan of nie ca be pte
nde sent mt oF Plt eg,
Oy met i DAY of am
level present, Fig 51.44) a ih conta image. Of
‘he 4 istprams shown, tbe ast histogram epee he
est imag,
(so in is chapter would be otras he Fes
tee histograms in te fourth histogram, Ia oer words,
‘we ty fincas he dynamic rang of he ing. Though
MATLAB code for iting he istogeam oan nag,
A Pop for bitgran
form=stia
italagn}==0 abe et
Wee ae Matin py ne opt ings
‘+ Minus oye ot inage
a0
s=2
Modes isogram ie
27
Grey [oi] a
Nember ot pies | 50 0 | 60
is and tes the
"epi image to occupy the eine
aay
cay
[ice aaa
sts Mier ater nar rchag
(ae NEN cig al wages ye mer sa
mec
= ete a he pete hig rein bart
oo, a iepret ctr sein
ar ah Let Whe anes example of hsp ina
“lense mcing
donne
Mee re? seo
a | anassateserm mt ne
6 Hiegran archon a,
| thy ys ccamaoe ate yh)oe) emg
9 Pciog haga othe anna! mag
| baat ny,
ges
g
3
ob BE
Carre
(eat
Fig S23 sae 90 ol
a aati
Tass pm orl ey
We se tat CDF ives us 8 at respon
rat te supe of he cipal histogram gS
Sota; ne doped he etait,
vain wing cotinous doin mae
For ine poseig We eed 0 werk inns
ori,
eae if i the total numberof pines in at
te he PDR
nae
Weknow,
se THe f ae
°
Inte discrete domain,
A TeeS pio. masiewec oF
o
Werake 1,2 andthe ls column
Otd ere evel
Equalized grey eel
We see tat 9) wich
pe
Mer rantomed on aia
sopra,
11,3.5,6.7. Thre a0 pie ine
Rune rye | Naber of ps
° °
1 1
2 °
3 1m
‘ °
5 ss
‘
2
So bere weave. A dat histogram becomes an evealy
‘pce histogram. Bat wat We had started ou o No etn Ss UUM ake ig
stew eat steady quid sagan t
sy ca we elie image 7
=
);
nurse
net eck)
Bch same a wha we bad gor ater
iain, eee
es
RE, equliing spun gives
alii
tialting so aay equalized is
te histogram. Go abet,
They wi es
8 the same reat ie,
inna id uty,
7 Youcan ry the program with "posi
adobe)
big=mesmasia)s
Gives maximum ral
lo colli)
(Crrontool 5% Gives the tl number
never 200);
ee 54 _ Additional Examples on
Histogram Modeting
Histogram of he rig image
HL
We conver the ciginat
presentation,
180200350
(Original bsogeam
NES © itay image Since We maxinui grey evel Is 15, we need 4:bits for its Fie.P.542()
tetsces teehee level rece, But one mga hing tt hype thatthe image comes much dak.
— 84d ‘The modified image is
=
J
Commence np er lye bea gy | zeae
nen hein py eed
a Histogram of the orignal and modified iomges is
Pate teen wim elie sownia Pig P4344)
- ™
Peo mu 34a he anni, we gs
5 orig try
‘uss a sam of histogram linear stretching. By
cy bles, we mean bislogram. From the image ita
2.6 fy and 27
12 8 4 ee
“Gray
Original histogram vel
te len
From the above formulation
m=z,
m3, ners
cen ign ei ip te WH GLE. BE py og
iether,
7 arama aaa SE
Sapa a ei pl
Fig P.54200)
acer
tess ce tote ey levels ec. Bu one nora hing tht happens i athe image becomes mich
exaas
ese i ope taoraos tlh inte. rae
egy eso Oil eg. Ae gg =O aad
2 [stale
sis[s[s
‘This i a sum of histogram linear seching. By
feomncy tube, we mean histogram. From the image i
eth = 2, ag = 6, = 804g =.
sige oats
awa
From the above formulation
When, mana,
med,
m4,
mes,
mes a7
Te modified images
an Prexine or image ehunceoea contin high
ee cov mt top epulzaton Wo where wpe
_-