6CS3- 01: Digital Image Pr ocessing

UNIT 1
Topic: Digital Image Repr esentation, Sampling & Quantization,
Steps in Image Pr ocessing, Image acquisition, Color Image
Repr esentation

Pr esented By: Yashika Saini,

Assistant Pr of essor Depar tment of
Computer Science Engineer ing Ar ya Institute
of Engineer ing & Technology, Jaipur
 Intr oduction to Image Pr ocessing
• Image is defined as a 2- dimensional
function, F(x, y), Wher e x & y ar e spatial
coor dinates, F is intensity or amplitude.
• An image is a pr oj ection of a 3D scene onto
a 2D pr oj ection plane.
• Types of Images:
1. Analog Image
2. Digital Image
• Analog Image: An analog image can be
r epr esented as a continous r ange of
values r epr esenting position (x, y) and
Intensity (F).
• Digital Image: When x, y & F ar e all finite
discr ete quantities, image is digital image.
• A digital image is composed of pictur
elements called e
pixels.
• Digital Image Pr ocessing: The analysis and
manipulation of a digitized image, in or der
to impr ove its quality is called digital
image pr ocessing.
 Applications of Digital Image Pr ocessing
1. Gamma- Ray Imaging: Maj or uses of imaging
based
on gamma r ays include nuclear medicine
and
astr onomical obser vations.
2. X- Ray Imaging:
• X- r ays ar e among the oldest sour ces of EM r adiation
used for imaging.
• X- r ays for medical and industr ial imaging ar e gener ated
using an X- r ay tube, which is a vacuum tube with a
cathode and anode
3. Imaging in the Ultr aviolet Band:
• They include lithogr aphy, industr ial
inspection,
micr oscopy, laser s, biological imaging,
and
astr onomical obser vations.
• Ultr aviolet light is used in f luor escence
micr oscopy,
one of the f astest gr owing ar eas of micr oscopy.
4. To impr ove Quality, r emove noise etc.

5. Imaging in the Micr owave Band:

• The pr incipal application of imaging in the
micr owave band is r adar .
• The unique f eatur e of imaging r adar is its ability
to collect data over vir tually any r egion at any
time, r egar dless of weather or ambient lighting
conditions.
Components of Image Pr ocessing
System
 Image Repr esentation
Bef or e we discuss image acquisition r ecall that a
digital image is composed of M r ows and N
columns of ro
pixels each stor ing a value. wf

Pixel values ar e most (r

o
of ten gr ey levels in the w,
co
r ange 0- 255(black- l)

white).
We will see later on
that images can
easily
be r epr esented as
matr icescol
 f (r ow, col)
r ow
 Image Sensing
• Incoming ener gy lands on a sensor
mater ial
r esponsive to that type of ener gy and this
gener ates a
voltage.
• Collections of sensor s ar e ar r anged to captur e
images.

1. Single Sensor
2. Line Sensor s 3.
Ar r ay Sensor s
 Image Acquisition
Images ar e typically gener ated by
illuminating a scene and absor bing the
ener gy r eflected by the obj ects in that
scene.
–Typical notions of
illumination and
scene
can be way of f :
• X- r ays of a skeleton
• Ultr asound of
an

unbor n baby
• Electr o-
micr oscopic

images of molecules
 Image Acquisition
Image Acquisition can be done in thr ee
ways: 1. Image Acquisition using Single
Sensor . 2. Image Acquisition using Sensor
Str ip.
3. Image Acquisition using sensor ar r ays.
 Images ar e analog
• Notice that we defined images as
functions in a continuous domain.
• Images ar e r epr esentations of an analog
wor ld.
• Hence, as with all digital signal
pr ocessing, we need to digitize our
images.
 Digitalization
• Digitalization of an analog signal
involves two oper ations:
• Sampling, and
• Quantization

• Both oper ations cor r espond to

discr etization of a quantity, but in
differ ent domains.
 Image Sampling & Quantization
• To conver t analog image into digital image we
have two inter mediate steps.
• Digitizing the coor dinates values (x,y) is called
sampling and digitizing the amplitude values (F) is
called quantization.
 Repr esenting Digital Images
• Let f (s,t) be a continuous image f unction, Wher e s
& t ar e continuous var iables.
• We conver t this f unction into a digital image by
sampling and quantization.
• Suppose that we sample the continuous image
into a 2- D ar r ay, f (x,y) , having M r ows and N
columns, wher e (x,y) ar e discr ete coor dinates.
• Wher e, x = 0, 1, 2,….., M - 1
y = 0, 1, 2,…..., N- 1.
 Repr esenting Digital Images
• Ther e ar e thr ee basic ways to r epr esent f(x,y).
 Digital Image Repr esentation
• Figur e shows a plot of the function, with two axes
deter mining spatial location and the thir d axis being the
values of f (intensities) as a f unction of the two spatial
var iables x and y.
• This type of r epr esentation is not pr efer able because in
dealing with complex images, inter polation becomes ver y
difficult.
Figur e: Image plotted as a sur f ace.
 Digital Image Repr esentation
• It shows f(x,y) as it would appear on a monitor or
photogr aph. Her e, the intensity of each point
is
pr opor tional to the value of f at that point.

Figur e: Image displayed as a visual intensity ar r ay

 Digital Image Repr esentation
• In this figur e, ther e ar e only thr ee discr ete
intensity values. If the intensity is
nor malized
to the inter val [0, 1], then each point in the
image has the value 0, 0.5, or 1.
• A monitor or pr inter simply conver ts these
thr ee values to black, gr ay, or
white,
r espectively.
 Digital Image Repr esentation
• The thir d r epr esentation is simply to display the
numer ical values of f(x,y) as an ar r ay (matr ix).
• When developing algor ithms, this r epr esentation is quite
useful when only par ts of the image ar e pr inted and
analyzed as numer ical values.

Figur e: 2- D numer ical ar r ay (0,

.5,
and 1) r epr esent black, gr ay,
and
white.
 Digital Image Repr esentation
• In equation for m, we wr ite the
r epr esentation of an M x N numer ical
ar r ay as

• Each element of this matr ix is called an

image element, pictur e element, pixel or
pel.
 Digital Image Repr esentation
• M & N should be positive integer s. But the
number of intensity levels L should
• Wher e, K is integer .
• Bits r equir ed to stor e a digitized image is
b=M x N x K
When M = N (No. of r ows= No. of column) ,
then Steps in Image Pr ocessing
 Image Acquisition
• The image is captur ed by a sensor (eg. Camer a),
and digitized if the output of the camer a or
sensor is not alr eady in digital f or m, using
analogue- to- digital conver tor .
 Image Enhancement
• The pr ocess of manipulating an image so that
the r esult is mor e suitable than the or iginal
f or specif ic applications.
• The idea behind enhancement techniques is
to br ing out details that ar e hidden, or
simple to highlight cer tain f eatur es of
inter est in an image.
Image Enhancement
 Image Restor ation
• Impr oving the appear ance of an image.

• Tend to be mathematical or
pr obabilistic
models. Enhancement, on the other hand,
is
based on human subj ective
pr efer ences
r egar ding what constitutes a
“good”
enhancement r esult.
 Image Restor ation
Color Image Pr ocessing
• Use the color of the image to extr act
featur es of inter est in an image.
• It includes color modeling and pr ocessing in
digital domain.
 Wavelets & Multi- Resolution
• Wavelets ar e small waves of limited
dur ation
which ar e used to calculate wavelet
tr ansfor m
which pr ovides time fr equency
infor mation.
• Wavelets lead to multir esolution
pr ocessing in
which images ar e r epr esented in
var ious
degr ees of r esolution.
 Compr ession
• Techniques for r educing the stor age
r equir ed to
save an image or the bandwidth r equir ed
to tr ansmit it.
 Mor phological Pr ocessing
• Tools for extr acting image components
that ar e useful in the r epr esentation and
descr iption of shape.

• In this step, ther e would be a tr ansition

fr om pr ocesses that output images, to
pr ocesses that output image attr ibutes.
 Image Segmentation
• Segmentation pr ocedur es par tition an
image into its constituent par ts or obj ects.
• The mor e accur ate the segmentation, the
mor e likely r ecognition is to succeed.
• It is gener ally used to locate obj ects and
boundar iesImage Segmentation
in obj ects.

Image Segmentation
Repr esentation and Descr iption
• Repr esentation: Make a decision whether
the
data should be r epr esented as a boundar y
or as a complete r egion. It is almost
always follows the output of a
segmentation stage.
• Boundar y Repr esentation: Focus on
exter nal shape char acter istics, such as
cor ner s and Inflections.
• Region Repr esentation: Focus on inter nal
pr oper ties, such as textur e or skeleton
Shape.
 Repr esentation and Descr iption
• Choosing a r epr esentation is only par t of the solution for
tr ansfor ming r aw data into a for m suitable for
subsequent
computer pr ocessing (mainly r ecognition).
• Descr iption: also called, featur e selection, deals
with
extr acting attr ibutes that r esult in some infor mation of
inter est.
 Obj ect Recognition
• Recognition: The pr ocess that assigns label
to
an obj ect based on the infor mation
pr ovided by its descr iption.
 Knowledge Base
• Knowledge about a pr oblem domain is
coded into an image pr ocessing system in
the for m of a knowledge database.
 Spatial and Intensity Resolution
• Spatial r esolution is a measur e of the smallest
discer nible details in an image.
• It can also be stated in number of ways, dots
(pixels) per unit distance or line pair s per unit
distance.
• Measur ing spatial r esolution
1. Dots per inch
2. Lines per
inch
3. Pixels per
inch
 Measur ing Spatial Resolution
1. Dots per inch
• DPI is a measur e of image r esolution.
• DPI means that how many dots of ink ar e pr inted per
inch. When an image get pr inted out fr om pr inter .
• The higher the DPI of the pr inter the higher is the
quality of the pr inted image or paper .
• Commonly used in pr inting and publishing industr y.
• For Newspaper s: 75dpi, 133dpi, Gloss
Magazines: y
Br ochur e:175dpi and Book pages: 2400
dpi.
 Measur ing Spatial Resolution
2. Lines per inch
• LPI is used to measur e the r esolution of images pr inted in
half tones (An image compr ised of such dots of one color is
usually called a half tone image). Because half tone images
ar e pr inted as a ser ies of dots.
• LPI is a measur ement of pr inting r esolution.
• LPI is usually used in laser pr inter s and gr aphic design.
 Measur ing Spatial Resolution
3. Pixels per inch
• PPI r efer s display r esolution, or how
many
individual pixels ar e displayed in one inch of a digital
image.
• PPI is measur e for differ ent devices such as tablets,
mobile
phones etc.
• The higher is the PPI, the higher is the quality.
 Intensity Resolution
• Intensity r esolution r efer s to the smallest
discer nible change in intensity level.
• Intensity of r esolution means the number
of pixels per squar e inch, which
deter mines the clar ity or shar pness of an
image.
 Illustr ation of the effects of r educing image spatial
r esolution
Effects of var ying the number of intensity levels in a digital
image Her e Value of K var ies fr om 8 to 1(8,7,6,5,4,3,2,1)
 Image Inter polation
• Inter polation is a tool which is used to r esize the
image such as zooming, shr inking, r otating and
geometr ic cor r ections.
• Image Inter polation is also called as r e- sampling
of image.
• In or der to r esize the image, we have to
r esample the image.
• For Example: Suppose that an image of size 500 ×
500 pixels has to be enlar ged 1.5 times to 750 ×
750 pixels.
 Near est Neighbor Inter polation
• In this method, it assigns to each new location the
intensity of its near est neighbor in the or iginal
image.

• Bilinear inter polation: In which we use the f our

near est neighbor s to estimate the intensity at a
given location.
• Bicubic inter polation: which involves the sixteen
near est neighbor s of a point .
 Intr oduction to Color Image Repr esentation

• Color is a power ful descr iptor which simplifies obj ect

identification and extr action fr om a scene.
• Human being can per ceive far mor e higher number of
color shades than gr ay scale shades.
• Color Image Pr ocessing is divided into two maj or
categor ies:
1. Full color : The images ar e acquir ed with a full- color
sensor , such as a color TV or color scanner .
2. Pseudo- color : Assign a color to a par ticular r ange of
intensities.
Pseudo color images ar e gr ayscale which ar e assigned
color based on the intensities values.
Full color & Pseudo- color
Pr ocessing
 Char acter ization of light
• If the light is Achr omatic, its only attr ibute
is its intensity. Achr omatic light is what
viewer s see on a black and white
television set.
• Chr omatic light spans the electr omagnetic
spectr um fr om 400 to 700nm.
• Thr ee descr ibe the o
quantities quality f
chr omatic light: r adiance luminance an
br ightness. , d
 Char acter ization of light
• Radiance (Watts- W): It is the total amount
of ener gy coming out of the light sour ce.
• Luminance (lumens- lm):It gives a measur e
of amount of ener gy an obser ver
per ceives fr om a light sour ce.
• Br ightness (no unit): It is subj ective
measur e that is pr actically impossible to
measur e. It cor r esponds to achr omatic
attr ibute of intensity.
 Color Standar dization
• CIE (Commission Inter national Eclair age)
has
standar dized specif ic wavelength values to
thr ee
pr imar y color s:
• Blue= 435.8nm
• Gr een= 546.1nm
• Red=700nm
 Pr imar y & Secondar y Color s
• The pr imar y color s can be added to pr oduce the
secondar y color s of light-
o Magenta (r ed plus blue)
o Cyan (gr een plus blue)
o Yellow (r ed plus gr een)

• Mixing the thr ee pr imar ies or secondar y with its

opposite pr imar y color , in the r ight intensities pr oduces
white light.
 Pr imar y & Secondar y Color s
• The pr imar y color s of pigments ar e magenta, cyan
and yellow and the secondar y color s ar e r ed, gr een
and blue.
• A pr oper combination of the thr ee pigment pr imar ies
or a secondar y with its opposite pr imar y pr oduces
black.
 Char acter izing Color
• One color can be distinguished f r om other by
using thr ee char acter istics.
• Br ightness- It embodies the achr omatic notion of
intensity.
• Hue- Dominant color (wavelength of light)
as
per ceived by an obser ver . It is measur e of color
of the obj ect & its hue is expr essed as an angle.
• Satur ation- It r ef er s to the r elative pur ity or the
amount of white light mixed with a hue. Or It is a
measur e of the r ichness of color .
 Assignment Questions
Q1 Def ine the image and digital image pr ocessing.
Explain the f undamental steps of digital
image
pr ocessing with suitable diagr am.
Q2 What ar e the applications of image pr ocessing?
Explain components of image pr ocessing system.
Q3 Explain image sensing & acquisition.
Q4 Explain image sampling and Quantization
pr ocess in
image pr ocessing?
Q5 Explain digital image r epr esentation.