What is Digital Image Processing?

An image is defined as a two-dimensional function,

f(x,y), where x and y are spatial (plane) coordinates, and
the amplitude of f at any pair of coordinates (x,y) is
called the intensity of the image at that point.
When x, y and the amplitude of f are all finite, discrete
quantities, we call the image a digital image.
The field of digital image processing refers to processing
digital images by using computers.
A digital image consists of a finite number of elements,
each of which has a particular location and value.
The elements are referred as picture elements, image
elements, pels and pixels.

Machines vs humans
Among the five senses, vision is
considered to be vital one for a human
being.
But a human being can perceive only
visible part of the electromagnetic
spectrum.
But machines can span the entire range of
electromagnetic spectrum from gamma to
radio waves.

Image vision vs Image analysis

Image processing is a branch in which
both the input and output of a process are
images.
The goal of computer vision is to use
computers to emulate human vision,
including learning, making inferences and
taking actions.
The area of image analysis is in between
image processing and computer vision.

Types of Processes in Image

Processing
There are 3 types of computerized processes.
They are : low, mid and high level processes.
Low-level processes involve primitive operations such as
image preprocessing to reduce noise, contrast
enhancement and image sharpening.
Here both the input and output are images.
Mid-level processing involves segmentation (partitioning
image into regions), description of objects to reduce
them to a form so that a computer can process and
classification (recognition) of objects.
Here inputs are images but outputs are attributes
extracted from images.
In high-level processing, we make sense of a collection
of recognized objects.

Example of dip and image analysis

The process of acquiring an image of a text,
processing it, extracting (segmenting) individual
characters, describing characters suitable for
computer processing and recognizing those
individual characters are in the scope of digital
image processing.
Making sense of the content of the page (text) is
viewed as the domain of image analysis and
computer vision.

Categorizing Images
One can categorize images according to their source.
One of the main sources is due to electromagnetic
spectrum.

The electromagnetic spectrum arranged according to energy per photon

The spectra band ranges from gamma rays (high

energy), X-rays, Ultraviolet, Visible, Infrared, Microwaves
to radio waves (low energy).
Other sources are acoustic, ultrasonic and electronic
(electron beams used in electron microscopy).

Gamma Ray Imaging

These are used in nuclear medicine and
astronomical observations.
In medicine, it is used for complete bone scan.
The nuclear imaging is also used in positron
emission tomography (PET).
It can render the 3-D image of patients.
It can detect tumors in the brain and lungs.
The images of stars which exploded about
15000 years ago, can be captured using gamma
rays.

Examples of Gamma Ray Images

(Bone scan and PET images)

X-ray Imaging
It is also used in medicine and astronomy.
We can get images of blood vessels in
angiography.
It is also used in Computerized Axial
Tomography (CAT) to generate 3-D
rendition of a patient.
High energy X-ray images are used in
industrial processes (electronic circuit
board).

Examples of X-ray Images

(Chest X-ray and Circuit boards)

Ultraviolet Imaging
It is used in lithography, industrial
inspection, microscopy, lasers, biological
imaging and astronomical observations.
Corn, cereals, onions caused by parasitic
fungi can be identified using this imaging
technique.

Examples of Ultraviolet Images

(Normal corn vs Smut corn)

Visible and Infrared Images

It is used in light microscopy, astronomy, remote
sensing, industry and law enforcement.
It is mainly used in weather observation and
prediction.
Visual inspection of manufactured goods use
visible spectrum.
Automated counting, license plate reading,
tracking and identifying bills, etc belong to visible
images.
Infrared images are used for night vision
systems.

Examples of visible and infrared images

(Microprocessor and surface of an audio CD)

Microwave Images
(Thumb print and paper currency)
Radar images belong to this category.
It can penetrate the inaccessible
regions of earths surface.

Radio Images
(MRI of human knee and spine)
It is used in Magnetic Resonance Imaging (MRI).

Acoustic imaging
The images use sound energy.
They are used in geological exploration,
industry and medicine.
It is also used in mineral and oil
exploration.
Ultrasonic images are used in obstetrics to
determine the health of unborn babies and
determining the sex of the baby.

Cross-sectional image of a seismic model

The arrow points to a hydrocarbon
(oil and/or gas) trap.

Fundamental steps in Digital

Image Processing

Fundamental steps of DIP

Image acquisition This stage involves preprocessing,

such as scaling.
Image enhancement Here we bring out details that
were obscured or highlight some features of interest in
an image. (eg) increasing the contrast of an image.
Image Restoration Here we talk about how to
improve the appearance of an image. Unlike
enhancement, which is subjective, this is objective.
Color Image Processing Due to Internet, this area is
becoming popular. Various color models are worthy to
know.
Wavelets Representing the images in various
degrees of resolution in the basis of wavelets.

Fundamental steps of DIP

Compression It is a technique for reducing the storage required

to save an image or bandwidth needed to transmit.
Morphological Processing It deals with tools for extracting
image components that are useful in the representation and
description of shape.
Segmentation These procedures partition an image into its
constituent parts or objects.
Representation and description It follows the output of a
segmentation stage.
It uses either the boundary of a region or all the points in the
region itself.
Description ( also called feature selection) deals with extracting
attributes or are basic for differentiating one class of objects from
another.
Recognition It is the process that assigns a label (eg. Vehicle) to
an object based on its descriptors.

Components of an Image
Processing System

Basic components of a general-purpose system

used for digital image processing

Image sensors Two elements are needed to

acquire digital images.
First one is the physical device that is sensitive to
energy radiated by the object that we want to
image.
The second one, called the digitizer, is a device for
converting the output of the physical sensing
device into digital form.
(eg) in a digital video camera, the sensors produce
an electrical output proportional to light intensity.
The digitizer converts these outputs to digital data.

Basic components of a general-purpose

system used for digital image processing

Specialized Image Processing Hardware It consists of digitizer plus hardware that performs
other primitive operations such as an arithmetic logic
unit (ALU), which performs arithmetic and logical
operations on entire image.
This type of hardware is also called as front-end
subsystem and its characteristic is speed.
This unit does things that require fast data throughputs
which main computer cannot handle.
Computer In an image processing system it is a
general-purpose computer.
Software It consists of specialized modules that does
specific tasks (eg. matlab)

Basic components of a general-purpose

system used for digital image processing

Mass storage An image of 1024 X 1024 size, storing the

intensity of each pixel in 8 bits, requires one megabyte of
storage.
For short-time storage, we can use computer memory.
Another method is to use a specialized board called frame
buffer, that store one or more images and can be accessed
rapidly.
They enable us to instantaneously zoom, scroll (vertical shift)
and pan (horizontal shift).
For on-line storage magnetic disks or optical-media are used.
The archival storage needs massive capacity but are accessed
infrequently.
Image Displays These are mainly color TV monitors.
Hardcopy These devices include laser printers, film cameras,
inkjet units, etc.

Brightness Adaptation
There are 2 phenomena which clearly
demonstrate that the perceived brightness
is not a simple function of intensity.
The visual system tends to undershoot or
overshoot around the boundary of regions
of different intensities.
These scalloped bands near the
boundaries are called Mach bands.

Image Formation Model

Image Formation Model

When an image is generated from a physical process, its
values are proportional to energy radiated by a physical
source (em waves).
Hence f(x,y) must be nonzero and finite. That is
0 f ( x, y )

The function f(x,y) is characterized by 2 components.

1) The amount of source illumination incident on the
scene being viewed called illumination component
denoted as i(x,y)
2) The amount of illumination reflected by the objects in
the scene called reflectance component denoted as
r(x,y).

Image Formation Model

f ( x , y ) i ( x, y ) r ( x , y )
where
and

0 f ( x, y )
0 r ( x, y ) 1

When r(x,y) is zero, we have total absorption and when it

is 1, we have total reflectance.
The nature of i(x,y) is determined by the illumination
source and r(x,y) is determined by the characteristics of
the imaged objects.
If the images are formed via transmission of illumination
through a medium such as a chest X-ray, we use
transmissivity instead of reflectivity function.

Illumination values of objects

On a clear day, sun produces about 90000
lm/m2 of illumination on the surface of the Earth.
It is about 10000 lm/ m2 on a cloudy day.
On a clear evening, a full moon gives about 0.1
lm/ m2 of illumination.
The typical illumination of an office is about 1000
lm/m2.
The typical values of r(x,y) are: 0.01 for black
velvet, 0.65 for stainless steel, 0.8 for white wall
paint, 0.90 for silver plated metal and 0.93 for
snow.

Image Sampling
The output of many sensors is continuous
voltage.
To get a digital image, we need to convert this
voltage into digital form.
But this involves 2 processes, namely sampling
and quantization.
An image is continuous with respect to x and y
coordinates and in amplitude.
Digitizing the coordinates is called sampling.
Digitizing the amplitude is called quantization.

Sampling and Quantization

Let us consider a gray scale image.
We can take intensity values along a particular line.
Subsequently, we consider few equally spaced points
(discrete locations) along this line and mark the intensity
values at these points called sampling points.
But the values of amplitude are continuous in nature.
The gray level values can also be converted (quantized)
into discrete quantities.
This is called quantization.
We have converted the gray level ranges into 4 levels.
For this we assign one of the 4 discrete gray levels
(closest one) to each sample.

Image to be sampled and quantized

along a scan line

The intensity variation sampled at

regular intervals along the scan line
300

250

200

150

100

50

10

20

30

40

50

60

70

80

90

100

Four level intensity quantization of

sampled scan line
260
240
220
200
180
160
140
120
100
80
60

10

20

30

40

50

60

70

80

90

100

Representing a Digital Image

Representing an image as a Matrix

The result of sampling and quantization is a
matrix of real numbers.
Thus a sampled digital image has M rows and N
columns.
The values of the coordinates (x,y) also has
become discrete due to quantization.
Thus we can write a M X N digital image as:

f(0,0) f(0,1)
f(0,N-1)

f(1,0) f(1,1)
f(1,N-1)
f(x,y) =
.
.
.

:
:
:

f(M-1,0) f(M-1,1) f(M-1,N-1)

Dynamic range of an image

The range of values spanned by the gray scale is
called dynamic range of an image.
An image with high dynamic range is said to be a
high contrast image.
An image with low dynamic range leads to a dull,
washed out look.
The number of bits required to store a digitized
image is

b M N k
When M = N, b = N^k
For example, a 256 X 256 image represented
with 8 bits takes 5,24,288 bits.

Dynamic Range of Images

Spatial and Gray-Level Resolution

Sampling determines the spatial resolution
of an image.
Resolution is the smallest number of
discernible line pairs per unit distance.
Gray-level resolution is the smallest
discernible change in gray level.

Image to be sub sampled

Size of the image is :512 X 512

Sub sampled images

Size of the image is :256 X 256

Size of the image is :128 X 128

Subsampling
The subsampling is achieved by deleting
appropriate number of rows and columns from
the original image.
For example, we can get a 256 X 256 image by
deleting the every other row and column from
the 512 X 512 image.
In order to see the effect of subsampling, we can
replicate appropriate rows and columns of 256 X
256 to bring it to 512 X 512 size.
We notice the checkboard pattern.

Size of the image is :512 X 512

Size of the image is :512 X 512

Size of the image is :512 X 512

Zooming and Shrinking Digital Images

Zooming is viewed as oversampling and shrinking is
viewed as under sampling.
Zooming is a 2 step process: the creation of new pixel
locations and assignment of gray levels to those new
locations.
For example, say we want to zoom an image of size 500
X 500 to 750 X 750.
We can use nearest neighbor interpolation for zooming.
Pixel replication is the special case of nearest neighbor
interpolation.
Pixel replication is used to zoom the image by an integer
number of times.
Here new locations are exact duplicates of old locations.
It is very fast but produces check board effect and hence
is undesirable for larger magnification.

Bilinear interpolation
The best way is to use bilinear interpolation
using four nearest neighbors of a point.
Let (x,y) denote the coordinates of a point in the
zoomed image and let v(x,y) denote the gray
level assigned to it.
The assigned gray level is given by

v(x,y) = ax + by + cxy + d
Here the four coefficients are determined from
the 4 equations in four unknowns that can be
written using the 4 nearest neighbors of point
(x,y). (Assignment)

Shrinking an image
For shrinking an image by one-half, we delete
every other row and column.
In order to shrink an image by non integer factor,
we expand the grid to fit over the original image,
do gray-level nearest neighbor or bilinear
interpolation, and then shrink the grid back to its
original specified size. (Assignment).
It is good to blur an image slightly before
shrinking it.

Image zooming using nearest

neighbor gray-level interpolation
Image size:128X128

Image size:256X256

Image size:512X512

Image Zooming using bilinear interpolation

Image size:128X128

Image size:256X256

Image size:512X512

Relationships between Pixels

Neighbors of a Pixel
A pixel p at coordinates (x,y) has 4 horizontal and
vertical neighbors whose coordinates are given by

(x+1,y), (x-1,y), (x,y+1), (x,y-1).

This set of pixels is called the 4-neighbors of p and
is denoted by N4(p). Some of the neighbors of p lie
outside the image if (x,y) is on the border of the
image.
The four diagonal neighbors of p have coordinates

(x+1, y+1), (x+1, y-1), (x-1, y+1), (x-1, y-1)

And are denoted by ND(p). These points, along with
4-neighbors are called 8-neighbors of p, denoted by
N8(p).

Adjacency and Connectivity

To establish if 2 pixels are connected, it must be
determined if they are neighbors and if their gray
levels satisfy a specified criterion of similarity.
(eg) in a binary image, 2 pixels may be 4
neighbors but they are said to be connected only
if they have the same value.
In a gray scale image, we consider any subset of
allowed gray level values for connectivity.

3 types of adjacency

4-adjacency: Two pixels p and q with values

from V are 4 adjacent if q is in the set N4(q).

8-adjacenty: Two pixels p and q with values

from V are 8-adjacent if q is in the set N8(q).

m-adjacency: Two pixels p and q with values

from V are m-adjacent if

q is in N4(p), or

q is in ND(p) and set N4(p) N4(q) has no pixels

whose values are from V.

Digital Path

Arrangement of pixels

pixels that are 8-adjacecnt (dashed)

A (digital) path (or curve) from pixel p with

coordinates (x,y) to pixel q with coordinates (s,t) is a
sequence of distinct pixels with coordinates
(x0 , y0), (x1 , y1 ), ,(xn , yn )
Here n is the length of the path.
If (x0 , y0) = (xn , yn ), the path is a closed path.

Distance Measures
For pixels p, q and z with coordinates (x,y), (s,t) and
(v,w) respectively, D is a distance function or metric if
D(p,q) 0
(D(p,q) = 0 iff p = q)
D(p,q) = D(q,p) and
D(p,z) D(p,q) + D(q,z)
The Euclidean distance between p and q is defined as

De(p,q) = [(x s)2 + (y t)2]1/2

Here the pixels having a distance less than or equal to
some value r from (x,y) are the points contained in a disk
of radius r centered at (x,y)

D4 distance
The D4 distance (called city-block distance) between p
and q is defined as

D4(p,q) = | x s| + | y t|.
Here the pixels have a D4 distance from (x,y) less than or
equal to some value r form a diamond centered at (x,y).
(eg) the pixels with D4 distance 2 from (x,y) (the center
point) form the following contours of constant distance:

2
1
2

2
1
0
1
2

2
1
2

2
The pixels with D4 = 1 are the 4-neighbors of (x,y).

D8 distance
The D8 distance (called chessboard distance) between p
and q is defined as

D8(p,q) = max(| x s|, |y t|).

The pixels with D8 distance from (x,y) less than or equal
to some value r form a square centered at (x,y).
(eg) the pixels with D8 distance 2 from (x,y) (center
point) form the following contours of constant distance:

2
2
2
2
2

2
1
1
1
2

2
1
0
1
2

2
1
1
1
2

2
2
2
2
2
The pixels with D8 = 1 are the 8-neighbors of (x,y).