lOMoARcPSD|49982872

ECE-VII- Image Processing Notes Main

Digital Image Processing (Visvesvaraya Technological University)

Scan to open on Studocu

Studocu is not sponsored or endorsed by any college or university

Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 1
IMAGE PROCESSING
Subject Code : 10EC763 IA Marks : 25
No. of Lecture Hrs/Week : 04 Exam Hours : 03
Total no. of Lecture Hrs. : 52 Exam Marks : 100
PART - A
UNIT - 1
DIGITAL IMAGE FUNDAMENTALS: What is Digital Image Processing.
fundamental Steps in Digital Image Processing, Components of an Image processing
system, elements of Visual Perception. 6 Hours
UNIT - 2
Image Sensing and Acquisition, Image Sampling and Quantization, SomeBasic
Relationships between Pixels, Linear and Nonlinear Operations. 6 Hours
UNIT - 3
IMAGE TRANSFORMS: Two-dimensional orthogonal & unitary transforms,
properties of unitary transforms, two dimensional discrete Fourier transform. 6 Hours
UNIT - 4
Discrete cosine transform, sine transform, Hadamard transform, Haar transform, Slant
transform, KL transform. 6 Hours
PART - B
UNIT - 5
IMAGE ENHANCEMENT: Image Enhancement in Spatial domain, SomeBasic Gray
Level Trans -formations, Histogram Processing, Enhancement Using Arithmetic/Logic
Operations. 6 Hours
UNIT - 6
Basics of Spatial Filtering Image enhancement in the Frequency Domain filters,
Smoothing Frequency Domain filters, Sharpening Frequency Domain filters,
homomorphic filtering. 6 Hours
UNIT - 7
Model of image degradation/restoration process, noise models, Restoration in the
Presence of Noise, Only-Spatial Filtering Periodic Noise Reduction by Frequency
Domain Filtering, Linear Position-Invariant Degradations, inverse filtering, minimum
mean square error (Weiner) Filtering 10 Hours
UNIT - 8
Color Fundamentals. Color Models, Pseudo color Image Processing., processing basics of
full color image processing 6 Hours

TEXT BOOK:
1. “Digital Image Processing”, Rafael C.Gonzalez and Richard E. Woods, Pearson
Education, 2001, 2nd edition.

REFERENCE BOOKS:
1. “Fundamentals of Digital Image Processing”, Anil K. Jain, Pearson Edun, 2001.
2. “Digital Image Processing and Analysis”, B. Chanda and D. Dutta Majumdar, PHI,
2003.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 2

INDEX SHEET
SL.NO TOPIC PAGE NO.
I Unit-1: DIGITAL IMAGE FUNDAMENTALS 1-8
1.1 What is Digital Image Processing?
1.2 fundamental Steps in Digital Image Processing,
1.3 Components of an Image processing system
1.4 elements of Visual Percepton
UNIT – 2 : DIGITAL IMAGE
II 9-18
FUNDAMENTALS
2.1 Image Sensing and Acquisition,
2.2 Image Sampling and Quantization,
2.3 SomeBasic Relationships between Pixels,
2.4 Linear and Nonlinear Operations
III UNIT - 3 : IMAGE TRANSFORMS 19-25
3.1 Two-dimensional orthogonal & unitary transforms
3.2 properties of unitary transforms,
two dimensional discrete Fourier
3.3
transform
IV UNIT – 4 : IMAGE TRANSFORMS 26-32
4.1 Discrete cosine transform
4.2 sine transform,
4.3 Hadamard transform
4.4 Haar transform,
4.5 Slant transform,
4.6 KL transform.
V UNIT – 5 : IMAGE ENHANCEMENT 33-41
5.1 Image Enhancement in Spatial domain,
5.2 Some Basic Gray Level Trans -formations
5.3 Histogram Processing, Enhancement
5.4 Using Arithmetic/Logic operations.
VI UNIT-6 IMAGE ENHANCEMENT 42-52
Basics of Spatial Filtering Image enhancement in the
6.2
Frequency Domain
6.3 filters,
6.4 Smoothing Frequency Domain filters,
Sharpening Frequency Domain
6.5
filters
6.6 homomorphic filtering
VII UNIT-7 IMAGE ENHANCEMENT 53-70
7.1 Model ofDownloaded
image degradation
by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 3
7.4 Restoration in the Presence of Noise,
Only-Spatial Filtering Periodic Noise Reduction by
7.5
Frequency Domain Filtering,
7.6 Linear Position-Invariant Degradations
7.7 inverse filtering
7.8 minimum mean square error (Weiner) Filtering
VIII UNIT-8 IMAGE ENHANCEMENT 71-87
Color Fundamentals.
8.5 Color Models,
8.6 Pseudo color Image Processing
processing basics of full color image processing

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 4

Image Processing 10EC763

Unit-1
Introduction

What Is Digital Image Processing?

An image may be 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 or gray level of the image at that
point. When x, y, and the amplitude values 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 means of a digital
computer. Note that a digital image is composed of a finite number of
elements, each of which has a particular location and value. These elements
are referred to as picture elements, image elements, pels, and pixels. Pixel is
the term most widely used to denote the elements of a digital image.

Fundamental Steps in Digital Image Processing

It is helpful to divide the material covered in the following chapters

into the two broad categories defined in Section 1.1: methods whose input
and output are images, and methods whose inputs may be images, but whose
outputs are attributes extracted from those images..The diagram does not
imply that every process is applied to an image. Rather, the intention is to
convey an idea of all the methodologies that can be applied to images for
different purposes and possibly with different objectives.
Image acquisition is the first process acquisition could be as simple as being
given an image that is already in digital form. Generally, the image
acquisition stage involves preprocessing, such as scaling.
Image enhancement is among the simplest and most appealing areas
of digital image processing. Basically, the idea behind enhancement
techniques is to bring out detail that is obscured, or simply to highlight
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 5

Image Processing 10EC763

certain features of interest in an image. A familiar example of enhancement is

when we increase the contrast of an image because “it looks better.” It is
important to keep in mind that enhancement is a very subjective area of
image processing

Image restoration is an area that also deals with improving the

appearance of an image. However, unlike enhancement, which is subjective,
image restoration is objective, in the sense that restoration techniques tend to
be based on mathematical or probabilistic models of image degradation.
Enhancement, on the other hand, is based on human subjective preferences
regarding what constitutes a “good” enhancement result.
Color image processing is an area that has been gaining in importance
because of the significant increase in the use of digital images over the
Internet. fundamental concepts in color models and basic color processing in

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 6

Image Processing 10EC763

a digital domain. Color is used also in later chapters as the basis for
extracting features of interest in an image.

Wavelets are the foundation for representing images in various

degrees of resolution. In particular, this material is used in this book for
image data compression and for pyramidal representation, in which images
are subdivided successively into smaller regions.
Compression, as the name implies, deals with techniques for
reducing the storage required to save an image, or the bandwidth required to
transmit it. Although storage technology has improved significantly over the
past decade, the same cannot be said for transmission capacity. This is true
particularly in uses of the Internet, which are characterized by significant
pictorial content. Image compression is familiar (perhaps inadvertently) to
most users of computers in the form of image file extensions, such as the jpg
file extension used in the JPEG(Joint Photographic Experts Group) image
compression standard.
Morphological processing deals with tools for extracting image
components that are useful in the representation and description of shape.
The material in this chapter begins a transition from processes that output
images to processes that output image attributes, Segmentation procedures
partition an image into its constituent parts or objects. In general, autonomous
segmentation is one of the most difficult tasks in digital image processing. A
rugged segmentation procedure brings the process a long way toward
successful solution of imaging problems that require objects to be identified
individually. On the other hand, weak or erratic segmentation algorithms
almost always guarantee eventual failure. In general, the more accurate the
segmentation, the more likely recognition is to succeed.
Representation and description almost always follow the output of a
segmentation stage, which usually is raw pixel data, constituting either the
boundary of a region (i.e., the set of pixels separating one image region from
another) or all the points in the region itself. In either case, converting the
data to aDownloaded
form suitable for
by Sudheer computer
Yadav processing is necessary. The first
([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 7

Image Processing 10EC763

decision that must be made is whether the data should be represented as a

boundary or as a complete region. Boundary representation is appropriate
when the focus is on external shape characteristics, such as corners and
inflections. Regional representation is appropriate when the focus is on
internal properties, such as texture or skeletal shape. In some applications,
these representations complement each other. Choosing a
representation is only part of the solution for transforming raw data into a
form suitable for subsequent computer processing. A method must also be
specified for describing the data so that features of interest are highlighted.
Description, also called feature selection, deals with extracting attributes that
result in some quantitative information of interest or are basic for
differentiating one class of objects from another.

Recognition is the process that assigns a label (e.g., “vehicle”) to an object

based on its descriptors. As detailed in Section 1.1, we conclude our coverage
of digital image processing with the development of methods for recognition
of individual objects. So far we have said nothing about the need for prior
knowledge or about the interaction between the knowledge base and
Knowledge about a problem domain is coded into an image processing
system in the form of a knowledge database. This knowledge may be as
simple as detailing regions of an image where the information of interest is
known to be located, thus limiting the search that has to be conducted in
seeking that information. The knowledge base also can be quite complex,
such as an interrelated list of all major possible defects in a materials
inspection problem or an image database containing high-resolution satellite
images of a region in connection with change-detection applications.

In addition to guiding the operation of each processing module, the

knowledge base also controls the interaction between modules. This
distinction is made in Fig. 1.23 by the use of double headed arrows between
the processing modules and the knowledge base, as opposed to single-headed
arrows linking the byprocessing
Downloaded modules.
Sudheer Yadav Although we do not discuss image
([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 8

Image Processing 10EC763

display explicitly at this point, it is important to keep in mind that viewing the
results of image processing can take place at the output of any stage.

Components of an Image Processing System

Although large-scale image processing systems still are being sold for
massive imaging applications, such as processing of satellite images, the
trend continues toward miniaturizing and blending of general-purpose small
computers with specialized image processing hardware.
The function of each component is discussed in the following
paragraphs, starting with image sensing. With reference to sensing, two
elements are required to acquire digital images. The first is a physical device
that is sensitive to the energy radiated by the object we wish to image. The
second, called a digitizer, is a device for converting the output of the physical
sensing device into digital form. For instance, in a digital video camera, the
sensors produce an electrical output proportional to light intensity. The
digitizer converts these outputs to digital data.
Specialized image processing hardware usually consists of the
digitizer just mentioned, plus hardware that performs other primitive
operations, such as an arithmetic logic unit (ALU), which performs
arithmetic and logical operations in parallel on entire images. One example of
how an ALU is used is in averaging images as quickly as they are digitized,
for the purpose of noise reduction. This type of hardware sometimes is called
a front-end subsystem, and its most
1.5

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 9

Image Processing 10EC763

distinguishing characteristic is speed. In other words, this unit performs

functions that require fast data throughputs (e.g., digitizing and averaging
video images at 30 frames_s) that the typical main computer cannot handle.
The computer in an image processing system is a general-purpose computer
and can range from a PC to a supercomputer. In dedicated applications,
sometimes specially designed computers are used to achieve a required level
of performance, but our interest here is on general-purpose image processing
systems. In these systems, almost any well-equipped PC-type machine is
suitable for offline image processing tasks.

Software for image processing consists of specialized modules that perform

specific tasks. A well-designed package also includes the capability for the
user to write code that, as a minimum, utilizes the specialized modules. More
sophisticated software packages allow the integration of those modules and
general- purpose software commands from at least one computer language.

Mass storage capability is a must in image processing applications.An image

of size 1024*1024 pixels, in which the intensity of each pixel is an 8-bit
quantity, requires one megabyte of storage space if the image is not
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 10

Image Processing 10EC763

compressed. When dealing with thousands, or even millions, of images,

providing adequate storage in an image processing system can be a challenge.
Digital storage for image processing applications falls into three principal
categories: (1) short term storage for use during processing, (2) on-line
storage for relatively fast recall, and (3) archival storage, characterized by
infrequent access. Storage is measured in bytes (eight bits), Kbytes (one
thousand bytes), Mbytes (one million bytes), Gbytes (meaning giga, or one
billion, bytes), and T bytes (meaning tera, or one trillion, bytes).

One method of providing short-term storage is computer memory.Another is

by specialized boards, called frame buffers, that store one or more images and
can be accessed rapidly, usually at video rates (e.g., at 30 complete images
per second).The latter method allows virtually instantaneous image zoom, as
well as scroll (vertical shifts) and pan (horizontal shifts). Frame buffers
usually are housed in the specialized image processing hardware unit. Online
storage generally takes the form of magnetic disks or optical-media storage.
The key factor characterizing on-line storage is frequent access to the stored
data. Finally, archival storage is characterized by massive storage
requirements but infrequent need for access. Magnetic tapes and optical disks
housed in “jukeboxes” are the usual media for archival applications.
Image displays in use today are mainly color (preferably flat screen) TV
monitors. Monitors are driven by the outputs of image and graphics display
cards that are an integral part of the computer system. Seldom are there
requirements for image display applications that cannot be met by display
cards available commercially as part of the computer system. In some cases,
it is necessary to have stereo displays, and these are implemented in the form
of headgear containing two small displays embedded in goggles worn by the
user.

Hardcopy devices for recording images include laser printers, film cameras,
heat-sensitive devices, inkjet units, and digital units, such as optical and CD-
ROM disks. Film provides
Downloaded the highest
by Sudheer Yadav possible resolution, but paper is the
([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 11

Image Processing 10EC763

obvious medium of choice for written material. For presentations, images are
displayed on film transparencies or in a digital medium if image projection
equipment is used. The latter approach is gaining acceptance as the standard
for image presentations.

Networking is almost a default function in any computer system in use today.

Because of the large amount of data inherent in image processing
applications, the key consideration in image transmission is bandwidth. In
dedicated networks, this typically is not a problem, but communications with
remote sites via the Internet are not always as efficient. Fortunately, this
situation is improving quickly as a result of optical fiber and other broadband
technologies.

Recommended Questions

1. What is digital image processing? Explain the fundamental steps in digital

image processing.
2. Briefly explain the components of an image processing system.
3. How is image formed in an eye? Explain with examples the perceived
brightness is not a simple function of intensity.
4. Explain the importance of brightness adaption and discrimination in image
processing.
5. Define spatial and gray level resolution. Briefly discuss the effects resulting
from a reduction in number of pixels and gray levels.
6. What are the elements of visual perception?

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 12

Image Processing 10EC763

UNIT – 2
Image Sensing and Acquisition,

The types of images in which we are interested are generated by the

combination of an “illumination” source and the reflection or absorption of
energy from that source by the elements of the “scene” being imaged. We
enclose illumination and scene in quotes to emphasize the fact that they are
considerably more general than the familiar situation in which a visible light
source illuminates a common everyday 3-D (three-dimensional) scene. For
example, the illumination may originate from a source of electromagnetic
energy such as radar, infrared, or X-ray energy. But, as noted earlier, it could
originate from less traditional sources, such as ultrasound or even a
computer-generated illumination pattern. Similarly, the scene elements could
be familiar objects, but they can just as easily be molecules, buried rock
formations, or a human brain. We could even image a source, such as
acquiring images of the sun.

Depending on the nature of the source, illumination energy is reflected from,

or transmitted through, objects. An example in the first category is light
reflected from a planar surface. An example in the second category is when
X-rays pass through a patient’s body for thepurpose of generating a
diagnostic X-ray film. In some applications, the reflected or transmitted
energy is focused onto a photo converter (e.g., a phosphor screen), which
converts the energy into visible light. Electron microscopy and some
applications of gamma imaging use this approach.
The idea is simple: Incoming energy is transformed into a voltage by the
combination of input electrical power and sensor material that is responsive
to the particular type of energy being detected.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 13

Image Processing 10EC763

The output voltage waveform is the response of the sensor(s), and a digital
quantity is obtained from each sensor by digitizing its response. In this
section, we look at the principal modalities for image sensing and generation.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 14

Image Processing 10EC763

Image Acquisition Using a Single Sensor

The components of a single sensor. Perhaps the most familiar sensor of this
type is the photodiode, which is constructed of silicon materials and whose
output voltage waveform is proportional to light. The use of a filter in front of
a sensor improves selectivity. For example, a green (pass) filter in front of a
light sensor favors light in the green band of the color spectrum. As a
consequence, the sensor output will be stronger for green light than for other
components in the visible spectrum. In order to generate a 2-D image using a
single sensor, there has to be relative displacements in both the x- and y-
directions between the sensor and the area to be imaged. Figure 2.13 shows
an arrangement used in high-precision scanning, where a film negative is
mounted onto a drum whose mechanical rotation provides displacement in
one dimension. The single sensor is mounted on a lead screw that provides
motion in the perpendicular direction. Since mechanical motion can be
controlled with high precision, this method is an inexpensive (but slow) way
to obtain high-resolution images. Other similar mechanical arrangements use
a flat bed, with the sensor moving in two linear directions. These types of
mechanical digitizers sometimes are referred to as microdensitometers.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 15

Image Processing 10EC763

Image Acquisition Using Sensor Strips

A geometry that is used much more frequently than single sensors consists of
an in-line arrangement of sensors in the form of a sensor strip, shows. The
strip provides imaging elements in one direction. Motion perpendicular to the
strip provides imaging in the other direction. This is the type of arrangement
used in most flat bed scanners. Sensing devices with 4000 or more in-line
sensors are possible. In-line sensors are used routinely in airborne imaging
applications, in which the imaging system is mounted on an aircraft that flies
at a constant altitude and speed over the geographical area to be imaged.
One-dimensional imaging sensor strips that respond to various bands of the
electromagnetic spectrum are mounted perpendicular to the direction of
flight. The imaging strip gives one line of an image at a time, and the motion
of the strip completes the other dimension of a two-dimensional image.
Lenses or other focusing schemes are used to project area to be scanned onto
the sensors.

Sensor strips mounted in a ring configuration are used in medical and

industrial imaging to obtain cross-sectional (“slice”) images of 3-D objects\

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 16

Image Processing 10EC763

Image Acquisition Using Sensor Arrays

The individual sensors arranged in the form of a 2-D array. Numerous

electromagnetic and some ultrasonic sensing devices frequently are arranged
in an array format. This is also the predominant arrangement found in digital
cameras. A typical sensor for these cameras is a CCD array, which can be
manufactured with a broad range of sensing properties and can be packaged
in rugged arrays of elements or more. CCD sensors are used widely in digital
cameras and other light sensing instruments. The response of each sensor is
proportional to the integral of the light energy projected onto the surface of
the sensor, a property that is used in astronomical and other applications
requiring low noise images. Noise reduction is achieved by letting the sensor
integrate the input light signal over minutes or even hours. The two
dimensional, its key advantage is that a complete image can be obtained by
focusing the energy pattern onto the surface of the array. Motion obviously is
not necessary, as is the case with the sensor arrangements

This figure shows the energy from an illumination source being

reflected from a scene element, but, as mentioned at the beginning of this
section, the energy also could be transmitted through the scene elements. The
first function performed by the imaging system is to collect the incoming
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 17

Image Processing 10EC763

energy and focus it onto an image plane. If the illumination is light, the front
end of the imaging system is a lens, which projects the viewed scene onto the
lens focal plane. The sensor array, which is coincident with the focal plane,
produces outputs proportional to the integral of the light received at each
sensor. Digital and analog circuitry sweep these outputs and convert them to
a video signal, which is then digitized by another section of the imaging
system.

Image Sampling and Quantization,

To create a digital image, we need to convert the continuous sensed
data into digital form. This involves two processes: sampling and
quantization. A continuous image, f(x, y), that we want to convert to digital
form. An image may be continuous with respect to the x- and y-coordinates,
and also in amplitude. To convert it to digital form, we have to sample the
function in both coordinates and in amplitude. Digitizing the coordinate
values is called sampling. Digitizing the amplitude values is called
quantization.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 18

Image Processing 10EC763

The one-dimensional function shown in Fig. 2.16(b) is a plot of amplitude

(gray level) values of the continuous image along the line segment AB. The
random variations are due to image noise. To sample this function, we take
equally spaced samples along line AB, The location of each sample is given
by a vertical tick mark in the bottom part of the figure. The samples are
shown as small white squares superimposed on the function. The set of these
discrete locations gives the sampled function. However, the values of the
samples still span (vertically) a continuous range of gray-level values. In
order to form a digital function, the gray-level values also must be converted
(quantized) into discrete quantities. The right side gray-level scale divided
into eight discrete levels, ranging from black to white. The vertical tick marks
indicate the specific value assigned to each of the eight gray levels. The
continuous gray levels are quantized simply by assigning one of the eight
discrete gray levels to each sample. The assignment is made depending on the
vertical proximity of a sample to a vertical tick mark. The digital samples
resulting from both sampling and quantization.

Some Basic Relationships Between Pixels

Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 19

Image Processing 10EC763

In this section, we consider several important relationships between

pixels in a digital image.As mentioned before, an image is denoted by f(x,
y).When referring in this section to a particular pixel, we use lowercase
letters, such as p and q.

Neighbors of a Pixel
A pixel p at coordinates (x, y) has four 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, called the 4-neighbors of p, is denoted by N4(p). Each
pixel is a unit distance from (x, y), and some of the neighbors of p lie outside
the digital 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, together with the 4-neighbors, are
called the 8-neighbors of p, denoted by N8(p). As before, some of the points
in ND(p) and N8(p) fall outside the image if (x, y) is on the border of the
image.

Adjacency, Connectivity, Regions, and Boundaries

Connectivity between pixels is a fundamental concept that simplifies

the definition of numerous digital image concepts, such as regions and
boundaries. To establish if two pixels are connected, it must be determined if
they are neighbors and if their gray levels satisfy a specified criterion of
similarity (say, if their gray levels are equal).For instance, in a binary image
with values 0 and 1, two pixels may be 4-neighbors, but they are said to be
connected only if they have the same value.
Let V be the set of gray-level values used to define adjacency. In a
binary image, V={1} if we are referring to adjacency of pixels with value 1.
In a grayscale image, the idea is the same, but set V typically contains more
elements. For example, in the adjacency of pixels with a range of possible
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 20

Image Processing 10EC763

gray-level values 0 to 255, set V could be any subset of these 256 values. We
consider three types of adjacency:

(a) 4-adjacency. Two pixels p and q with values from V are 4-adjacent if q is
in the set N4(p).
(b) 8-adjacency. Two pixels p and q with values from V are 8-adjacent if q is
in the set N8(p).2.5
(c) m-adjacency (mixed adjacency).Two pixels p and q with values from V
are m-adjacent if
(i) q is in N4(p), or
(ii) q is in ND(p) and the set has no pixels whose values are from V.

Linear and Nonlinear Operations

Let H be an operator whose input and output are images. H is said to be a
linear operator if, for any two images f and g and any two scalars a and b,
H(af + bg) = aH(f) + bH(g).

In other words, the result of applying a linear operator to the sum of two
images (that have been multiplied by the constants shown) is identical to
applying the operator to the images individually, multiplying the results by
the appropriate constants, and then adding those results. For example, an
operator whose function is to compute the sum of K images is a linear
operator. An operator that computes the absolute value of the difference of
two images is not.
Linear operations are exceptionally important in image processing because
they are based on a significant body of well-understood theoretical and
practical results. Although nonlinear operations sometimes offer better
performance, they are not always predictable, and for the most part are not
well understood theoretically.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 21

Image Processing 10EC763

Recommended Questions
1. Explain the concept of sampling and quantization of an image.
2. Explain i) false contouring ii) checkboard pattern
3. How image is acquired using a single sensor? Discuss.
4. Explain zooming and shrinking digital images.
5. Define 4-adjacency, 8 – adjacency and m – adjacency.
6. With a suitable diagram, explain how an image is acquired using a circular
sensor strip.
7. Explain the relationships between pixels . and also the image operations on a
pixel basis.
8. Explain linear and nonlinear operations.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 22

Image Processing 10EC763

Unit-3
UNITARY TRANSFORMS
One dimensional signals

For a one dimensional sequence { f ( x), 0 x N 1} represented as a vector

T
f f (0) f (1)  f ( N 1) of size N , a transformation may be written as
N 1
g T f g (u) T (u, x) f ( x), 0 u N 1
x 0

where g (u) is the transform (or transformation) of f (x) , and T (u, x) is the so
called

Forward transformation kernel. Similarly, the inverse transform is the

relation
N 1
f ( x) I ( x, u) g (u), 0 x N 1
u 0
or written in a matrix form
1
f I g T g
where I ( x, u) is the so called inverse transformation kernel.

If
1 T
I T T

the matrix T is called unitary, and the transformation is called unitary as

well. It can be proven (how?) that the columns (or rows) of an N N unitary
matrix are orthonormal and therefore, form a complete set of basis vectors in
the N dimensional vector space.
In that case
T N 1
f T g f ( x) T (u, x) g (u )
u 0
T T
The columns of T , that is, the vectors T u T (u,0) T (u,1) T (u, N 1)
are called the basis vectors of T .

Two dimensional signals (images)

As a one dimensional signal can be represented by an orthonormal set of

basis vectors, an image can also be expanded in terms of a discrete set of
basis arrays called basis images through a two dimensional (image)
transform.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 23

Image Processing 10EC763

For an N N image f ( x, y ) the forward and inverse transforms are given

below
N 1N 1
g (u, v) T (u, v, x, y) f ( x, y)
x 0 y 0
N 1N 1
f ( x, y) I ( x, y, u, v) g (u, v)
u 0v 0
where, again, T (u, v, x, y) and I ( x, y, u, v) are called the forward and inverse
transformation kernels, respectively.

The forward kernel is said to be separable if

T (u, v, x, y) T1 (u, x)T2 (v, y)

It is said to be symmetric if T1 is functionally equal to T2 such that

T (u, v, x, y) T1 (u, x)T1 (v, y)

The same comments are valid for the inverse kernel.

If the kernel T (u, v, x, y) of an image transform is separable and symmetric,
N 1N 1 N 1N 1
then the transform g (u, v) T (u, v, x, y) f ( x, y) T1 (u, x)T1 (v, y) f ( x, y)
x 0 y 0 x 0 y 0

can be written in matrix form as follows

T
g T1 f T1
where f is the original image of size N N , and T 1 is an N N
transformation matrix with elements tij T1 (i, j ) . If, in addition, T 1 is a
unitary matrix then the transform is called separable unitary and the original
image is recovered through the relationship

T
f T1 g T1

Fundamental properties of unitary transforms

The property of energy preservation

In the unitary transformation

g T f
1 T
it is easily proven (try the proof by using the relation T T ) that
2 2
g f

Thus, a unitary transformation preserves the signal energy. This property is

called energy preservation property.
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 24

Image Processing 10EC763

This means that every unitary transformation is simply a rotation of the

vector f in the N - dimensional vector space.

For the 2-D case the energy preservation property is written as

N 1N 1 2 N 1N 1 2
f ( x, y ) g( u, v )
x 0y 0 u 0v 0

The property of energy compaction

Most unitary transforms pack a large fraction of the energy of the image into
relatively few of the transform coefficients. This means that relatively few of
the transform coefficients have significant values and these are the coefficients
that are close to the origin (small index coefficients).
This property is very useful for compression purposes. (Why?)

THE TWO DIMENSIONAL FOURIER TRANSFORM

Continuous space and continuous frequency

The Fourier transform is extended to a function f ( x, y ) of two variables. If

f ( x, y ) is continuous and integrable and F(u, v ) is integrable, the following
Fourier transform pair exists:
j 2 ( ux vy)
F (u , v) f ( x, y )e dxdy

1
f ( x, y ) F (u , v)e j 2 ( ux vy)
dudv
(2 ) 2
In general F(u, v ) is a complex-valued function of two real frequency
variables u, v and hence, it can be written as:
F (u, v) R(u, v) jI (u, v)
The amplitude spectrum, phase spectrum and power spectrum, respectively,
are defined as follows.
F (u, v ) R2 (u, v ) I 2 (u, v )
1 I (u , v )
(u , v ) tan
R (u , v )
2
P(u, v ) F(u, v ) R2 (u, v ) I 2 (u, v )

2.2 Discrete space and continuous frequency

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 25

Image Processing 10EC763

For the case of a discrete sequence f ( x, y ) of infinite duration we can define

the 2-D discrete space Fourier transform pair as follows

j ( xu vy)
F ( u, v ) f ( x, y ) e
x y

1
f ( x, y ) F (u, v )e j ( xu vy )
dudv
(2 ) 2 u v

F(u, v ) is again a complex-valued function of two real frequency variables

u, v and it is

periodic with a period 2 2 , that is to say

F(u, v )F(u 2 , v ) F(u, v 2 )

The Fourier transform of f ( x, y ) is said to converge uniformly when F(u, v )
is finite and
N1 N2
j ( xu vy )
lim lim f ( x, y )e F (u, v ) for all u, v .
N1 N2 x N1 y N2

When the Fourier transform of f ( x, y ) converges uniformly, F(u, v ) is an

analytic function and is infinitely differentiable with respect to u and v .

Discrete space and discrete frequency: The two dimensional Discrete Fourier
Transform (2-D DFT)

If f ( x, y ) is an M N array, such as that obtained by sampling a continuous

function of two dimensions at dimensions M and N on a rectangular grid,
then its two dimensional Discrete Fourier transform (DFT) is the array given
by

1 M 1N 1
j 2 ( ux / M vy / N )
F (u, v) f ( x, y)e
MN x 0 y 0

u 0,, M 1 , v 0,, N 1

and the inverse DFT (IDFT) is

M 1N 1
f ( x, y ) F ( u, v )e j 2 ( ux / M vy / N )

u 0 v 0

When images are sampled in a square array, M N and

1 N 1N 1
j 2 ( ux vy )/ N
F ( u, v ) f ( x, y )e
N x 0y 0

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 26

Image Processing 10EC763

1 N 1N 1
f ( x, y ) F ( u, v )e j 2 ( ux vy )/ N
N u 0v 0

It is straightforward to prove that the two dimensional Discrete Fourier

Transform is separable, symmetric and unitary.

Properties of the 2-D DFT

Most of them are straightforward extensions of the properties of the 1-D

Fourier Transform. Advise any introductory book on Image Processing.

The importance of the phase in 2-D DFT. Image reconstruction from amplitude
or phase only.

The Fourier transform of a sequence is, in general, complex-valued, and the

unique representation of a sequence in the Fourier transform domain requires
both the phase and the magnitude of the Fourier transform. In various
contexts it is often desirable to reconstruct a signal from only partial domain
information. Consider a 2-D sequence f ( x, y ) with Fourier transform
F ( u, v ) f ( x, y ) so that

j (u , v )
F (u, v) { f ( x, y} F (u, v) e f

It has been observed that a straightforward signal synthesis from the Fourier
transform phase f (u , v ) alone, often captures most of the intelligibility of the

original image f ( x, y ) (why?). A straightforward synthesis from the Fourier

transform magnitude F (u , v) alone, however, does not generally capture the
original signal’s intelligibility. The above observation is valid for a large
number of signals (or images). To illustrate this, we can synthesise the phase-
only signal f p ( x, y ) and the magnitude-only signal f m ( x, y) by

1 j (u , v )
f p ( x, y) 1e f

1
f m ( x, y ) F (u , v) e j 0

and observe the two results (Try this exercise in MATLAB).

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 27

Image Processing 10EC763

An experiment which more dramatically illustrates the observation that

phase-only signal synthesis captures more of the signal intelligibility than
magnitude-only synthesis, can be performed as follows.
Consider two images f ( x, y ) and g ( x, y) . From these two images, we
synthesise two other images f1 ( x, y ) and g1 ( x, y) by mixing the amplitudes
and phases of the original images as follows:
1 j (u , v )
f1 ( x, y) G(u, v) e f

1 j (u , v )
g1 ( x, y) F (u, v) e g
In this experiment f1 ( x, y ) captures the intelligibility of f ( x, y ) , while
g1 ( x, y) captures the intelligibility of g ( x, y) (Try this exercise in
MATLAB).

THE DISCRETE COSINE TRANSFORM (DCT)

One dimensional signals

This is a transform that is similar to the Fourier transform in the sense that the
new independent variable represents again frequency. The DCT is defined
below.
N 1 (2 x 1)u
C (u ) a(u ) f ( x) cos , u 0,1,, N 1
x 0 2N
with a(u ) a parameter that is defined below.
1/ N u 0
a (u )
2/ N u 1,, N 1
The inverse DCT (IDCT) is defined below.
N 1 (2 x 1)u
f ( x) a(u )C (u ) cos
u 0 2N

Two dimensional signals (images)

For 2-D signals it is defined as

N 1N 1 (2 x 1)u (2 y 1)v
C (u, v) a(u )a(v) f ( x, y ) cos cos
x 0 y 0 2N 2N
N 1N 1 (2 x 1)u (2 y 1)v
f ( x, y ) a(u )a(v)C (u, v) cos cos
u 0v 0 2N 2N
a(u ) is defined as above and u, v 0,1, , N 1

Properties of the DCT transform

Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 28

Image Processing 10EC763

 The DCT is a real transform. This property makes it attractive in comparison

to the Fourier transform.
 The DCT has excellent energy compaction properties. For that reason it is
widely used in image compression standards (as for example JPEG
standards).
 There are fast algorithms to compute the DCT, similar to the FFT for
computing the DFT.

Recommended Questions

1. Define two-dimensional DFT. Explain the following properties of 2-DFT.

i) Translation ii) rotation iii) distributivity and scaling iv) separability
2. What are basis vectors?
3. Derive the expression for 2D circular convolution theorem.
4. Define two – dimensional unitary transform. Check whether the unitary DFT
matrix is unitary or not for N = 4.
5. For the 2 X 2 transform A and the image U
1 1 1 2
A = 1/1 1 -1 and U = 8 4

Calculate the transformed image V and the basis images.

6. Consider the image segment shown in fig
i) Let V = {0, 1}. Compute the lengths of shortest 4 - , 8 – and m – paths between p
and q.
ii) Repeat for V = {1, 2}.
3 1 2 1 (q)
3 2 0 2
( p) 1 2 1 1

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 29

Image Processing 10EC763

UNIT – 4

WALSH TRANSFORM (WT)

One dimensional signals

This transform is slightly different from the transforms you have met so far.
Suppose we have a function f ( x), x 0,, N 1 where N 2 n and its Walsh
transform W (u) .
If we use binary representation for the values of the independent variables x
and u we
need n bits to represent them. Hence, for the binary representation of x and
u we can

write: ( x)10 bn 1 ( x)bn 2 ( x) b0 ( x) 2 , (u )10 bn 1 (u )bn 2 (u ) b0 (u ) 2

with bi (x) 0 or 1 for i 0,, n 1 .

Example

If f ( x), x 0,,7, (8 samples) then n 3 and for x 6 ,

6 = (110)2 b2 (6) 1,b1 (6) 1,b0 (6) 0

We define now the 1-D Walsh transform as

1 N 1 n 1
W (u ) f ( x) ( 1) bi ( x )bn 1 i (u ) or
N x 0 i 0
n 1
bi ( x ) bn 1 i (u )
1 N 1
W (u ) f ( x)( 1) i 0

N x 0
The array formed by the Walsh kernels is again a symmetric matrix having
orthogonal rows and columns. Therefore, the Walsh transform is and its
n 1
elements are of the form T (u, x) ( 1)bi ( x )bn 1 i (u ) . You can immediately
i 0

observe that T (u, x) 0 or 1 depending on the values of bi (x) and bn 1 i (u ) . If

the Walsh transform is written in a matrix form
W T f
for u 0 we see that (u )10 bn 1 (u )bn 2 (u )b0 (u ) 2
000 2 and hence,
1 N 1
bn 1 i (u ) 0 , for any i . Thus, T (0, x) 1 and W (0) f ( x) . We see that
N x 0

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 30

Image Processing 10EC763

the first element of the Walsh transform in the mean of the original function
f (x) (the DC value) as it is the case with the Fourier transform.

The inverse Walsh transform is defined as follows.

N 1 n 1
f ( x) W (u ) ( 1) bi ( x )bn 1 i (u ) or
u 0 i 0
n 1
N 1 bi ( x ) bn 1 i (u )
f ( x) W (u)( 1) i 0

u 0

Two dimensional signals

The Walsh transform is defined as follows for two dimensional signals.

1 N 1N 1 n 1
W ( u, v ) f ( x, y ) ( 1) ( bi ( x )bn 1 i ( u ) bi ( y ) bn 1 i ( v ))
or
N x 0 y 0 i 0
n 1
(bi ( x ) bn 1 i (u) bi ( y ) bn 1 i ( v ))
1 N 1N 1
W ( u, v ) f ( x, y )( 1) i 0

N x 0 y 0

The inverse Walsh transform is defined as follows for two dimensional

signals.
1 N 1N 1 n 1
f ( x, y ) W ( u, v ) ( 1) ( bi ( x )bn 1 i ( u ) bi ( y ) bn 1 i ( v ))
or
N u 0v 0 i 0
n 1
(bi ( x ) bn 1 i (u) bi ( y ) bn 1 i ( v ))
1 N 1N 1
f ( x, y ) W (u, v )( 1) i 0
N u 0v 0

4.3 Properties of the Walsh Transform

 Unlike the Fourier transform, which is based on trigonometric terms, the

Walsh transform consists of a series expansion of basis functions whose
values are only 1 or 1 and they have the form of square waves. These
functions can be implemented more efficiently in a digital environment than
the exponential basis functions of the Fourier transform.
 The forward and inverse Walsh kernels are identical except for a constant
1
multiplicative factor of for 1-D signals.
N
 The forward and inverse Walsh kernels are identical for 2-D signals. This is
because the array formed by the kernels is a symmetric matrix having
orthogonal rows and columns, so its inverse array is the same as the array
itself. Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 31

Image Processing 10EC763

 The concept of frequency exists also in Walsh transform basis functions. We

can think of frequency as the number of zero crossings or the number of
transitions in a basis vector and we call this number sequency. The Walsh
transform exhibits the property of energy compaction as all the transforms
that we are currently studying. (why?)
 For the fast computation of the Walsh transform there exists an algorithm
called Fast Walsh Transform (FWT). This is a straightforward
modification of the FFT. Advise any introductory book for your own interest.

HADAMARD TRANSFORM (HT)

Definition

In a similar form as the Walsh transform, the 2-D Hadamard transform is defined as
follows.

Forward
1 N 1N 1 n 1
H ( u, v ) f ( x, y ) ( 1) ( bi ( x )bi ( u ) bi ( y ) bi ( v ))
, N 2n or
N x 0 y 0 i 0
n 1
(bi ( x ) bi ( u ) bi ( y ) bi ( v ))
1 N 1N 1
H ( u, v ) f ( x, y )( 1) i 0
N x 0 y 0

Inverse
1 N 1N 1 n 1
f ( x, y ) H ( u, v ) ( 1) ( bi ( x )bi ( u ) bi ( y ) bi ( v ))
etc.
N u 0v 0 i 0

Properties of the Hadamard Transform

 Most of the comments made for Walsh transform are valid here.
 The Hadamard transform differs from the Walsh transform only in the order
of basis functions. The order of basis functions of the Hadamard transform
does not allow the fast computation of it by using a straightforward
modification of the FFT. An extended version of the Hadamard transform is
the Ordered Hadamard Transform for which a fast algorithm called Fast
Hadamard Transform (FHT) can be applied.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 32

Image Processing 10EC763

 An important property of Hadamard transform is that, letting H N represent

the matrix of order N , the recursive relationship is given by the expression
HN HN
H 2N
HN HN

KARHUNEN-LOEVE (KLT) or HOTELLING TRANSFORM

The Karhunen-Loeve Transform or KLT was originally introduced as a series

expansion for continuous random processes by Karhunen and Loeve. For
discrete signals Hotelling first studied what was called a method of principal
components, which is the discrete equivalent of the KL series expansion.
Consequently, the KL transform is also called the Hotelling transform or the
method of principal components. The term KLT is the most widely used.

The case of many realisations of a signal or image (Gonzalez/Woods)

The concepts of eigenvalue and eigevector are necessary to understand the KL

transform.

If C is a matrix of dimension n n , then a scalar is called an eigenvalue of C if

n
there is a nonzero vector e in R such that
Ce e
The vector e is called an eigenvector of the matrix C corresponding to the
eigenvalue .

(If you have difficulties with the above concepts consult any elementary
linear algebra book.)

Consider a population of random vectors of the form

x1
x2
x

xn
The mean vector of the population is defined as
mx E x
The operator E refers to the expected value of the population calculated
theoretically using the probability density functions (pdf) of the elements x i
and the joint probability density functions between the elements x i and x j .
The covariance matrix of the population is defined as
T
Downloaded by SudheerCYadav
x ( x m x )( x m x )
E([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 33

Image Processing 10EC763

Because x is n -dimensional, C x and ( x m x )( x m x )T are matrices of order

n n . The element c ii of C x is the variance of x i , and the element c ij of C x
is the covariance between the elements x i and x j . If the elements x i and x j
are uncorrelated, their covariance is zero and, therefore, cij c ji 0.
For M vectors from a random population, where M is large enough, the
mean vector and covariance matrix can be approximately calculated from the
vectors by using the following relationships where all the expected values are
approximated by summations
1 M
mx xk
M k 1
1 M
T T
Cx xk xk mx mx
M k 1

Very easily it can be seen that C x is real and symmetric. In that case a set of
n orthonormal (at this point you are familiar with that term) eigenvectors
always exists. Let e i and i , i 1,2,, n , be this set of eigenvectors and
corresponding eigenvalues of C x , arranged in descending order so that

i i 1 for i 1,2,, n 1. Let A be a matrix whose rows are formed from

the eigenvectors of C x , ordered so that the first row of A is the eigenvector
corresponding to the largest eigenvalue, and the last row the eigenvector
corresponding to the smallest eigenvalue.
Suppose that A is a transformation matrix that maps the vectors x' s into
vectors y' s by using the following transformation
y A( x m x )
The above transform is called the Karhunen-Loeve or Hotelling transform.
The mean of the y vectors resulting from the above transformation is zero
(try to prove that)
my 0

the covariance matrix is (try to prove that)

T
Cy AC x A
and C y is a diagonal matrix whose elements along the main diagonal are the
eigenvalues of C x (try to prove that)

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 34

Image Processing 10EC763

1 0
2
Cy

0 n

The off-diagonal elements of the covariance matrix are 0 , so the elements of

the y vectors are uncorrelated.

Lets try to reconstruct any of the original vectors x from its corresponding
y . Because the rows of A are orthonormal vectors (why?), then A 1 AT ,
and any vector x can by recovered from its corresponding vector y by using
the relation
T
x A y mx
Suppose that instead of using all the eigenvectors of C x we form matrix AK
from the K eigenvectors corresponding to the K largest eigenvalues,
yielding a transformation matrix of order K n . The y vectors would then be
K dimensional, and the reconstruction of any of the original vectors would
be approximated by the following relationship
T
x̂ AK y mx

The mean square error between the perfect reconstruction x and the
approximate reconstruction x̂ is given by the expression
n K n
ems j j j .
j 1 j 1 j K 1

By using AK instead of A for the KL transform we achieve compression of

the available data.

The case of one realisation of a signal or image

The derivation of the KLT for the case of one image realisation assumes that
the two dimensional signal (image) is ergodic. This assumption allows us to
calculate the statistics of the image using only one realisation. Usually we
divide the image into blocks and we apply the KLT in each block. This is
reasonable because the 2-D field is likely to be ergodic within a small block
since the nature of the signal changes within the whole image. Let’s suppose
that f is a vector obtained by lexicographic ordering of the pixels f ( x, y)

within a block of size M M (placing the rows of the block sequentially).

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 35

Image Processing 10EC763

The mean vector of the random field inside the block is a scalar that is
estimated by the approximate relationship
1 M2
mf f (k )
M2 k 1

and the covariance matrix of the 2-D random field inside the block is C f
where
1 M2
cii f (k ) f (k ) m 2f
M2 k 1
and
1 M2
cij ci j f (k ) f (k i j ) m2f
M2 k 1
After knowing how to calculate the matrix C f , the KLT for the case of a
single realisation is the same as described above.

6.3 Properties of the Karhunen-Loeve transform

Despite its favourable theoretical properties, the KLT is not used in practice
for the following reasons.
 Its basis functions depend on the covariance matrix of the image, and hence
they have to recomputed and transmitted for every image.
 Perfect decorrelation is not possible, since images can rarely be modelled as
realisations of ergodic fields.
 There are no fast computational algorithms for its implementation.

Recommended Questions

1. Construct Haar transform matrix for N = 2.

2. Explain the importance of discrete cosine transform, with its properties.
3. Define DCT and its inverse transformation .
4. Discuss any three properties of discrete cosine transform.
5. Develop Hadamard transform for n = 3.
6. Discuss the properties of the Hadamard transform .
7. Derive the relation between DCT and DFT.
8. Write H matrix for the Haar transform for N = 8 and explain how it is
constructed.
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 36

Image Processing 10EC763

UNIT – 5

Preliminaries

Spatial domain methods

Suppose we have a digital image which can be represented by a two

dimensional random field f ( x, y ) .

An image processing operator in the spatial domain may be expressed as a

mathematical function T applied to the image f ( x, y ) to produce a new
image g ( x, y) T f ( x, y) as follows.
g ( x, y) T f ( x, y)

The operator T applied on f ( x, y ) may be defined over:

(i) A single pixel ( x, y ) . In this case T is a grey level transformation (or mapping)
function.
(ii) Some neighbourhood of ( x, y ) .
(iii) T may operate to a set of input images instead of a single image.

Example 1
The result of the transformation shown in the figure below is to produce an
image of higher contrast than the original, by darkening the levels below m
and brightening the levels above m in the original image. This technique is
known as contrast stretching.

s T (r )

r
m
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 37

Image Processing 10EC763

Example 2

The result of the transformation shown in the figure below is to produce a

binary image.
s T (r )

r
m
Frequency domain methods

Let g( x, y) be a desired image formed by the convolution of an image

f ( x, y ) and a linear, position invariant operator h( x, y ) , that is:
g ( x, y ) h( x, y ) f ( x, y )
The following frequency relationship holds:
G(u, v) H (u, v) F (u, v)
We can select H (u, v ) so that the desired image
1
g ( x, y ) H ( u, v ) F ( u, v )
exhibits some highlighted features of f ( x, y ) . For instance, edges in f ( x, y )
can be accentuated by using a function H (u, v ) that emphasises the high
frequency components of F (u, v) .

Spatial domain: Enhancement by point processing

We are dealing now with image processing methods that are based only on
the intensity of single pixels.

Intensity transformations

Image Negatives
The negative of a digital image is obtained by the transformation function
s T (r) L 1 r shown in the following figure, where L is the number of

grey levels. The idea is that the intensity of the output image decreases as the
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 38

Image Processing 10EC763

intensity of the input increases. This is useful in numerous applications such

as displaying medical images.

s T (r )

L 1

L 1 r
Contrast Stretching

contrast images occur often due to poor or non uniform lighting conditions,
or due to nonlinearity, or small dynamic range of the imaging sensor. In the
figure of Example 1 above you have seen a typical contrast stretching
transformation.

Histogram processing. Definition of the histogram of an image.

By processing (modifying) the histogram of an image we can create a new

image with specific desired properties.
Suppose we have a digital image of size N N with grey levels in the range
[0, L 1] . The histogram of the image is defined as the following discrete
function:
nk
p( rk )
N2
Where

rk is the kth grey level, k 0,1,, L 1

n k is the number of pixels in the image with grey level rk
N 2 is the total number of pixels in the image
The histogram represents the frequency of occurrence of the various grey
levels in the image. A plot of this function for all values of k provides a
global description of the appearance of the image.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 39

Image Processing 10EC763

Question: Think how the histogram of a dark image, a bright image and
an image of very low contrast would like. Plot its form in each case.

Global histogram equalisation

In this section we will assume that the image to be processed has a

continuous intensity that lies within the interval [0, L 1] . Suppose we divide
the image intensity with its maximum value L 1 . Let the variable r
represent the new grey levels (image intensity) in the image, where now
0 r 1 and let p r (r ) denote the probability density function (pdf) of the
variable r . We now apply the following transformation function to the
intensity
r
s T (r) pr ( w)dw , 0 r 1
0
(1) By observing the transformation of equation (1) we immediately see that
it possesses the following properties:

(i) 0 s 1 .
(ii) r2 r1 T (r2 ) T (r1 ) , i.e., the function T (r ) is increase ng with r .
0 1
(iii) s T (0) pr ( w)dw 0 and s T (1) p r ( w)dw 1 . Moreover, if the original
0 0
image has intensities only within a certain range [rmin , rmax] then
rmin rmax
s T (rmin ) pr ( w)dw 0 and s T (rmax) pr ( w)dw 1 since
0 0
pr (r ) 0, r rmin and r rmax . Therefore, the new intensity s takes always all
values within the available range [0 1].

Suppose that Pr (r ) , Ps (s) are the probability distribution functions (PDF’s)

of the variables r and s respectively.

Let us assume that the original intensity lies within the values r and r dr
with dr a small quantity. dr can be assumed small enough so as to be able to
consider the function pr (w) constant within the interval [r , r dr ] and equal
to pr (r ) . Therefore,
r dr r dr
Pr [r , r dr ] pr ( w)dw pr (r ) dw pr (r )dr .
r r
Now suppose that s T (r ) and s1 T (r dr ) . The quantity dr can be
assumed small enough so as to be able to consider that s1 s ds with ds

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 40

Image Processing 10EC763

small enough so as to be able to consider the function ps (w) constant within

the interval [s, s ds ] and equal to ps (s ) . Therefore,
s ds s ds
Ps [ s, s ds ] ps ( w)dw ps ( s) dw ps ( s)ds
s s
Since s T (r ) , s ds T (r dr ) and the function of equation (1) is increasing
with r , all and only the values within the interval [r , r dr ] will be mapped
within the interval [s, s ds ] . Therefore,
1
r T (s)
dr
Pr [r , r dr ] Ps [s, s ds] pr (r )dr ps ( s )ds ps ( s ) pr ( r )
ds r T 1
(s)

From equation (1) we see that

ds
pr (r )
dr
and hence,
1
ps ( s ) pr (r ) 1, 0 s 1
pr (r ) r T 1
(s)

Conclusion
From the above analysis it is obvious that the transformation of equation (1)
converts the original image into a new image with uniform probability
density function. This means that in the new image all intensities are present
[look at property (iii) above] and with equal probabilities. The whole range of
intensities from the absolute black to the absolute white are explored and the
new image will definitely have higher contrast compared to the original
image.

Unfortunately, in a real life scenario we must deal with digital images. The
discrete form of histogram equalisation is given by the relation
k nj k
sk T (rk ) 2
pr (rj ), 0 rk 1, k 0,1,, L 1
j 0N j 0

(2) The quantities in equation (2) have been defined in Section 2.2. To see
results of histogram equalisation look at any introductory book on Image
Processing.
The improvement over the original image is quite evident after using the
technique of histogram equalisation. The new histogram is not flat because
of the discrete approximation of the probability density function with the
histogram function. Note, however, that the grey levels of an image that has
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 41

Image Processing 10EC763

been subjected to histogram equalisation are spread out and always reach
white. This process increases the dynamic range of grey levels and produces
an increase in image contrast.

Local histogram equalisation

Global histogram equalisation is suitable for overall enhancement. It is often

necessary to enhance details over small areas. The number of pixels in these

areas my have negligible influence on the computation of a global

transformation, so the use of this type of transformation does not necessarily

guarantee the desired local enhancement. The solution is to devise

transformation functions based on the grey level distribution – or other

properties – in the neighbourhood of every pixel in the image. The histogram

processing technique previously described is easily adaptable to local

enhancement. The procedure is to define a square or rectangular

neighbourhood and move the centre of this area from pixel to pixel. At each

location the histogram of the points in the neighbourhood is computed and a

histogram equalisation transformation function is obtained. This function is

finally used to map the grey level of the pixel centred in the neighbourhood.

The centre of the neighbourhood region is then moved to an adjacent pixel

location and the procedure is repeated. Since only one new row or column of

the neighbourhood changes during a pixel-to-pixel translation of the region,

updating the histogram obtained in the previous location with the new data

introduced at each motion step is possible quite easily. This approach has

obvious advantages over repeatedly computing the histogram over all pixels
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 42

Image Processing 10EC763

in the neighbourhood region each time the region is moved one pixel

location. Another approach often used to reduce computation is to utilise non

overlapping regions, but this methods usually produces an undesirable

checkerboard effect.

Histogram specification

Suppose we want to specify a particular histogram shape (not necessarily

uniform) which is capable of highlighting certain grey levels in the image.
Let us suppose that:
p r (r ) is the original probability density function

p z (z ) is the desired probability density function

Suppose that histogram equalisation is first applied on the original image r
r
s T (r) pr ( w)dw
0
Suppose that the desired image z is available and histogram equalisation is
applied as well
z
v G( z ) p z ( w)dw
0
p s (s ) and p v (v ) are both uniform densities and they can be considered as

identical. Note that the final result of histogram equalisation is independent

z
of the density inside the integral. So in equation v G ( z ) p z ( w)dw we can
0

use the symbol s instead of v .

The inverse process z G 1 ( s ) will have the desired probability density
function. Therefore, the process of histogram specification can be
summarised in the following steps.
(i) We take the original image and equalise its intensity using the relation
r
s T (r) pr ( w)dw .
0
(ii) From the given probability density function p z (z ) we specify the probability
distribution function G(z ) .
(iii) We apply the inverse transformation function
1 1
z G ( s) G T (r)

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 43

Image Processing 10EC763

Spatial domain: Enhancement in the case of many realisations

of an image of interest available
Image averaging

Suppose that we have an image f ( x, y) of size M N pixels corrupted by

noise n( x, y) , so we obtain a noisy image as follows.

g ( x, y ) f ( x, y ) n( x, y )

For the noise process n( x, y) the following assumptions are made.

(i) The noise process n( x, y)
is ergodic.
1 M 1N 1
(ii) It is zero mean, i.e., E n( x, y ) n ( x, y ) 0
MN x 0 y 0

(ii) It is white, i.e., the autocorrelation function of the noise process

defined as
1 M 1 kN 1 l
R[k , l ] E{n( x, y)n( x k , y l )} n( x, y)n( x k , y l ) is
(M k )( N l ) x 0 y 0
zero, apart for the pair [k , l ] [0,0] .
Therefore,
1 M 1 kN 1 l
2 2
R k, l n( x, y )n( x k , y l ) n( x, y ) (k , l ) where n( x, y )
( M k )( N l ) x 0 y 0
is the variance of noise.

Suppose now that we have L different noisy realisations of the same image
f ( x, y) as gi ( x, y ) f ( x, y ) ni ( x, y ) , i 0,1,, L . Each noise process ni ( x, y )
satisfies the properties (i)-(iii) given above. Moreover, n2i ( x, y ) 2
. We form
the image g ( x, y) by averaging these L noisy images as follows:
1 L 1 L 1 L
g ( x, y) gi ( x, y) ( f ( x, y) ni ( x, y)) f ( x, y) ni ( x, y)
Li 1 Li 1 Li 1

Therefore, the new image is again a noisy realisation of the original image
1 L
f ( x, y) with noise n( x, y) ni ( x, y) .
Li 1

The mean value of the noise n( x, y) is found below.

1 L 1 L
E{n( x, y)} E{ ni ( x, y)} E{ni ( x, y)} 0
Li 1 Li 1

The variance of the noise n( x, y) is now found below.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 44

Image Processing 10EC763

2 2
2 1 L 1 L
n( x, y ) E{n 2 ( x, y )} E ni ( x, y ) E ni ( x, y )
Li 1 L2 i 1

1 L
2 1 L L 1 L
2 1 L L
E{( ni ( x , y ) )} E{( ( ni ( x, y )n j ( x, y ))} E{ni ( x, y )} E{ ni ( x, y )n j ( x, y )}
L2 i 1 L2 i 1j 1 L2 i 1 L2 i 1j 1
i j i j

1 L
2 1 2
0
L2 i 1 L

Therefore, we have shown that image averaging produces an image g ( x, y) ,

corrupted by noise with variance less than the variance of the noise of the
2
original noisy images. Note that if L we have n( x, y ) 0 , the resulting

noise is negligible.

Recommended Questions

1. What is the importance of image enhancement in image processing? Explain

in brief any two point processing techniques implemented in image
processing.
2. Explain histogram equalization technique.
3. What is histogram matching? Explain the development and implementation
of the method.
4. Highlight the importance of histograms in image processing and develop a
procedure to perform histogram equalization.
5. Explain the following image enhancement techniques, highlighting their area
of application.
i) Intensity level slicing
ii) Power – law transformation
6. Explain the following image enhancement techniques, highlighting their area
of application.
i) Bit – plane slicing.
ii) AND and OR operation

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 45

Image Processing 10EC763

UNIT - 6

patial domain: Enhancement in the case of a single image

Spatial masks

Many image enhancement techniques are based on spatial operations

performed on local neighbourhoods of input pixels. The image is usually
convolved with a finite impulse response filter called spatial mask. The use of
spatial masks on a digital image is called spatial filtering. Suppose that we
have an image f ( x, y ) of size N 2 and we define a neighbourhood around
each pixel. For example let this neighbourhood to be a rectangular window of
size 3 3

If we replace each pixel by a weighted average of its neighbourhood pixels

9
then the response of the linear mask for the pixel z 5 is wi zi . We may
i 1
repeat the same process for the whole image.

4.2 Lowpass and highpass spatial filtering

A 3 3 spatial mask operating on an image can produce (a) a smoothed

version of the image (which contains the low frequencies) or (b) it can
enhance the edges and suppress essentially the constant background
information. The behaviour is basically dictated by the signs of the elements
of the mask.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 46

Image Processing 10EC763

Let us suppose that the mask has the following form

To be able to estimate the effects of the above mask with relation to the sign
of the coefficients a, b, c, d , e, f , g, h , we will consider the equivalent one
dimensional mask

Let us suppose that the above mask is applied to a signal x(n) . The output of
this operation will be a signal y(n) as
y (n ) dx (n 1) x (n ) ex (n 1) Y (z ) dz 1 X ( z) X ( z) ezX ( z )
1 Y ( z) 1
Y ( z) ( dz 1 ez ) X ( z ) H ( z) dz 1 ez .
X ( z)

This is the transfer function of a system that produces the above input-output
relationship. In the frequency domain we have
H (e j ) d exp( j ) 1 e exp( j ) .

The values of this transfer function at frequencies 0 and are:

H (e j ) d 1 e
0

H (e j ) d 1 e

If a lowpass filtering (smoothing) effect is required then the following

condition must hold
H (e j ) H (e j ) d e 0
0

If a highpass filtering effect is required then

H (e j ) H (e j ) d e 0
0

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 47

Image Processing 10EC763

The most popular masks for lowpass filtering are masks with all their
coefficients positive and for highpass filtering, masks where the central pixel
is positive and the surrounding pixels are negative or the other way round.

Popular techniques for lowpass spatial filtering

Uniform filtering

The most popular masks for lowpass filtering are masks with all their
coefficients positive and equal to each other as for example the mask shown
below. Moreover, they sum up to 1 in order to maintain the mean of the
image.

1 1 1

1
9
1 1 1

1 1 1

Gaussian filtering

The two dimensional Gaussian mask has values that attempts to approximate
the continuous function
x2 y2
1 2
G ( x, y ) 2
e
2
In theory, the Gaussian distribution is non-zero everywhere, which would
require an infinitely large convolution kernel, but in practice it is effectively
zero more than about three standard deviations from the mean, and so we can
truncate the kernel at this point. The following shows a suitable integer-
valued convolution kernel that approximates a Gaussian with a of 1.0.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 48

Image Processing 10EC763

1 4 7 4 1

4 16 26 16 4

1
273 7 26 41 26 7

4 16 26 16 4

1 4 7 4 1

Median filtering

The median m of a set of values is the value that possesses the property that
half the values in the set are less than m and half are greater than m . Median
filtering is the operation that replaces each pixel by the median of the grey
level in the neighbourhood of that pixel.

Median filters are non linear filters because for two sequences x(n) and
y (n) median x (n ) y n
( ) median x (n ) median y(n)

Median filters are useful for removing isolated lines or points (pixels) while
preserving spatial resolutions. They perform very well on images containing
binary (salt and pepper) noise but perform poorly when the noise is
Gaussian. Their performance is also poor when the number of noise pixels in
the window is greater than or half the number of pixels in the window (why?)

Isolated
0 0 0 0 0 0
Median filtering
0 1 0 0 0 0

0 0 0 0 0 0

Directional smoothing

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 49

Image Processing 10EC763

To protect the edges from blurring while smoothing, a directional averaging

filter can be useful. Spatial averages g ( x, y : ) are calculated in several
selected directions (for example could be horizontal, vertical, main diagonals)
1
g ( x, y : ) f (x k, y l)
N ( k ,l ) W

and a direction is found such that f ( x, y ) g ( x, y : ) is minimum. (Note

that W is the neighbourhood along the direction and N is the number of
pixels within this neighbourhood). Then by replacing
g ( x, y : ) with g ( x, y : ) we get the desired result.

High Boost Filtering

A high pass filtered image may be computed as the difference between the
original image and a lowpass filtered version of that image as follows:
(Highpass part of image) = (Original) - (Lowpass part of image)
Multiplying the original image by an amplification factor denoted by A ,
yields the so called high boost filter:

(Highboost image) = ( A) (Original)-(Lowpass) = ( A 1) (Original)+(Original)-

(Lowpass) = ( A 1) (Original) + (Highpass)

The general process of subtracting a blurred image from an original as given

in the first line is called unsharp masking. A possible mask that implements
the above procedure could be the one illustrated below.

0 0 0 -1 -1 -1
1
0 A 0 9
-1 -1 -1

0 0 0 -1 -1 -1

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 50

Image Processing 10EC763

-1 -1 -1

1
-1 9A 1 -1
9

-1 -1 -1

The high-boost filtered image looks more like the original with a degree of
edge enhancement, depending on the value of A .

Popular techniques for highpass spatial filtering. Edge detection using

derivative filters

About two dimensional high pass spatial filters

An edge is the boundary between two regions with relatively distinct grey
level properties. The idea underlying most edge detection techniques is the
computation of a local derivative operator. The magnitude of the first
derivative calculated within a neighbourhood around the pixel of interest, can
be used to detect the presence of an edge in an image.
The gradient of an image f ( x, y) at location ( x, y) is a vector that consists of
the partial
derivatives of f ( x, y) as follows.
f ( x, y )
x
f ( x, y )
f ( x, y )
y

The magnitude of this vector, generally referred to simply as the gradient f

is
2 2 1/ 2
f ( x. y ) f ( x, y )
f ( x, y) mag( f ( x, y))
x y
Common practice is to approximate the gradient with absolute values which
is simpler to implement as follows.
f ( x, y ) f ( x, y )
f ( x, y )
x y
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 51

Image Processing 10EC763

(1) Consider a pixel of interest f ( x, y) z5 and a rectangular

neighbourhood of size
3 3 9 pixels (including the pixel of interest) as shown below.
y

z1 z2 z3

z4 z5 z6

z7 z8 z9

Roberts operator
Equation (1) can be approximated at point z 5 in a number of ways. The
simplest is to use the difference ( z5 z8 ) in the x direction and ( z5 z6 ) in
the y direction. This approximation is known as the Roberts operator, and is
expressed mathematically as follows.
f z5 z8 z5 z6

(2) Another approach for approximating (1) is to use cross differences

f z5 z9 z6 z8
(3) Equations (2), (3) can be implemented by using the following masks. The
original image is convolved with both masks separately and the absolute
values of the two outputs of the convolutions are added.

1 0 1 -1

-1 0 0 0

Roberts operator

1 0 0 1

0 -1 -1 0

Roberts operator

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 52

Image Processing 10EC763

Prewitt operator
Another approximation of equation (1) but using now a 3 3 mask is the
following.

f ( z7 z8 z 9 ) ( z1 z2 z3 ) ( z3 z6 z 9 ) ( z1 z4 z7 )
(4)
This approximation is known as the Prewitt operator. Equation (4) can be
implemented by using the following masks. Again, the original image is
convolved with both masks separately and the absolute values of the two
outputs of the convolutions are added.

Sobel operator.

Definition and comparison with the Prewitt operator

The most popular approximation of equation (1) but using a 3 3 mask is the
following.

f ( z7 2 z8 z9 ) ( z1 2 z2 z3 ) ( z3 2 z6 z9 ) ( z1 2 z4 z7 )
(5)

This approximation is known as the Sobel operator.

y
-1 0 1 -1 -1 -1

-1 0 1 0 0 0

-1 0 1 1 1 1

Prewitt operator
x

y
-1 0 1 -1 -2 -1

-2 0 2 0 0 0

-1 0 1 1 2 1

Sobel operator
Downloadedxby Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 53

Image Processing 10EC763

If we consider the left mask of the Sobel operator, this causes differentiation
along the y
direction. A question that arises is the following: What is the effect caused by
the same
mask along the x direction?

If we isolate the following part of the mask

1

and treat it as a one dimensional mask, we are interested in finding the effects
of that mask. We will therefore, treat this mask as a one dimensional impulse
response h[n] of the form h[n]

1 n 1 2
2 n 0
or h[n]
1 n 1 1
0 otherwise

-1 0 1 n

The above response applied to a signal x[n] yields a signal

y[n] x[n 1] 2 x[n] x[n 1] or in z-transform domain
1
Y ( z) (z 2 z) X ( z) Y( j ) 2(cos 1) X ( j ) . Therefore,

h[n] is the impulse response of a system with transfer function

H( j ) 2(cos 1) H ( j ) shown in the figure below for [0, ] . This is a

lowpass filter type of response. Therefore, we can claim that the Sobel
operator has a differentiation effect along one of the two directions and a
smoothing effect along the other direction.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 54

Image Processing 10EC763

3.5

2.5

1.5

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5

The same analysis for the Prewitt operator would give

1
Y ( z) (z 1 z) X ( z)
Y ( j ) (2 cos 1) X ( j ) H( j ) 2 cos 1

shown in the figure below for [0, ] . This response looks “strange” since it
decreases up to the point 2 cos 1 0 cos 0.5 and then starts
increasing.

2.5

1.5

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5

Based on the above analysis it is stated in the literature that the Sobel
operator have the advantage of providing both a differencing a smoothing
effect while Prewitt does not. However, if you implement both operators you
cannot see any visual difference.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 55

Image Processing 10EC763

Laplacian operator

The Laplacian of a 2-D function f ( x, y ) is a second order derivative defined

as

2 2
2 f ( x, y ) f ( x, y )
f ( x, y )
x2 y2

In practice it can be also implemented using a 3x3 mask as follows (why?)

2
f 4 z5 ( z2 z4 z6 z8 )

The main disadvantage of the Laplacian operator is that it produces double

edges (why?).

Recommended Questions

1. Explain the smoothing of images in frequency domain using:

i) Ideal low pass filter
ii) Butterworth lowpass filter
2. With a block diagram and equations, explain the homomorphic filtering. How
dynamic range compression and contrast enhancement is simultaneously
achieved?
3. Discuss homomorphic filtering.
4. Explain sharpening filters in the frequency domain
5. Explain the basic concept of spatial filtering in image enhancement and hence
explain the importance of smoothing filters and median filters.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 56

Image Processing 10EC763

Unit-7

Preliminaries
What is image restoration?

Image Restoration refers to a class of methods that aim to remove or reduce

the degradations that have occurred while the digital image was being
obtained. All natural images when displayed have gone through some sort of
degradation:
during display mode
during acquisition mode, or
during processing mode
The degradations may be due to
sensor noise
blur due to camera misfocus
relative object-camera motion
random atmospheric turbulence
others
In most of the existing image restoration methods we assume that the
degradation process can be described using a mathematical model.

How well can we do?

Depends on how much we know about

the original image
the degradations
(how accurate our models are)

Image restoration and image enhancement differences

Image restoration differs from image enhancement in that the latter is

concerned more with accentuation or extraction of image features rather than
restorationDownloaded
of degradations.
by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 57

Image Processing 10EC763

Image restoration problems can be quantified precisely, whereas

enhancement criteria are difficult to represent mathematically.
Image observation models
Typical parts of an imaging system: image formation system, a detector and a
recorder. A general model for such a system could be:
y(i, j ) r w(i, j ) n(i, j )

w(i, j) H f (i, j) h(i, j, i , j ) f (i , j )di dj

n(i, j ) g r[w(i, j )] n1 (i, j ) n2 (i, j )

Where y(i, j ) is the degraded image, f (i, j ) is the original image and
h(i, j, i , j ) is an operator that represents the degradation process, for example

a blurring process. Functions g and r are generally nonlinear, and

represent the characteristics of detector/recording mechanisms. n(i, j ) is the
additive noise, which has an image-dependent random component
g r H [ f (i, j )] n1 (i, j ) and an image-independent random component n2 (i, j ) .

Detector and recorder models

The response of image detectors and recorders in general is nonlinear. An

example is the response of image scanners
r (i , j ) w(i, j )
where and are device-dependent constants and w(i, j ) is the input
blurred image.
For photofilms
r(i, j ) log 10 w(i, j ) r0
[

where is called the gamma of the film, w(i, j ) is the incident light intensity
and r (i, j ) is called the optical density. A film is called positive if it has
negative .

Noise models

The general noise model

Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 58

Image Processing 10EC763

n(i, j ) g r[w(i, j )] n1 (i, j ) n2 (i, j )

is applicable in many situations. Example, in photoelectronic systems we

may have
g ( x) x . Therefore,
n(i, j ) w(i, j ) n1 (i, j ) n2 (i, j )
where n1 and n 2 are zero-mean, mutually independent, Gaussian white noise
fields. The term n 2 (i, j ) may be referred as thermal noise. In the case of films
there is no thermal noise and the noise model is

n (i , j ) log 10 w(i, j ) ro n1 (i, j )

Because of the signal-dependent term in the noise model, restoration

algorithms are quite
difficult. Often w(i, j ) is replaced by its spatial average, w , giving

n(i, j ) gr w n1 (i, j ) n2 (i, j )

which makes n(i, j ) a Gaussian white noise random field. A lineal

observation model for
photoelectronic devices is y (i, j ) w(i, j ) w n1 (i, j ) n2 (i, j )

For photographic films with 1

y (i, j ) log 10 w(i, j ) r0 an1 ( x, y )
where r0 , a are constants and r0 can be ignored.

The light intensity associated with the observed optical density y(i, j ) is

y (i, j ) an1 ( i , j )
I (i, j ) 10 w(i, j )10 w(i, j )n(i, j )
where n(i, j ) ˆ 10 an1 ( i , j )
now appears as multiplicative noise having a log-
normal distribution.

Keep in mind that we are just referring to the most popular image
observation models. In the literature you can find a quite large number
of different image observation models. Image restoration algorithms are
based on the above image formation models.

A general model of a simplified digital image degradation process

Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 59

Image Processing 10EC763

A simplified version for the image restoration process model is

y (i, j ) H f (i, j ) n(i, j )

Where y(i, j ) the degraded image

f (i, j ) the original image

H an operator that represents the degradation process
n(i, j ) the external noise which is assumed to be image-
independent

Possible classification of restoration methods

Restoration methods could be classified as follows:

deterministic: we work with sample by sample processing of the observed
(degraded) image
stochastic : we work with the statistics of the images involved
in the process
non-blind : the degradation process H is known
blind : the degradation process H is unknown
semi-blind : the degradation process H could be considered partly
known

From the viewpoint of implementation:

direct
iterative
recursive

Linear position invariant degradation models

Definition
We again consider the general degradation model
y (i, j ) H f (i, j ) n(i, j )

If we ignore the presence of the external noise n(i, j ) we get

y (i, j ) H f (i, j )
H is linear if
H k1 f1 (i, j ) k2 f 2 (i, j ) k1H f1 (i, j ) k2 H f 2 (i, j )

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 60

Image Processing 10EC763

H is position (or space) invariant if

H f (i a, j b) y(i a, j b)
From now on we will deal with linear, space invariant type of degradations.

In a real life problem many types of degradations can be approximated

by linear, position invariant processes.

Advantage: Extensive tools of linear system theory become available.

Disadvantage: In some real life problems nonlinear and space variant

models would be more appropriate for the description of the degradation
phenomenon.

Typical linear position invariant degradation models

Motion blur. It occurs when there is relative motion between the object and
the camera during exposure.
1 L L
, if i
h (i ) L 2 2
0, otherwise
Atmospheric turbulence. It is due to random variations in the reflective
index of the medium between the object and the imaging system and it occurs
in the imaging of astronomical objects.
i2 j2
h(i, j ) K exp
2 2
Uniform out of focus blur
1
, if i 2 j 2 R
h (i , j ) R
0, otherwise
Uniform 2-D blur
1 L L
2
, if i, j
h (i , j ) ( L) 2 2
0, otherwise

Some characteristic metrics for degradation models

Blurred Signal-to-Noise Ratio (BSNR): a metric that describes the

degradation model.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 61

Image Processing 10EC763

1 2
g ( i , j ) g (i , j )
MN i j
BSNR 10log 10 2
n

g (i, j ) y(i, j ) n(i, j )

g (i, j ) E{g (i, j )}

[
2
n : variance of additive noise
Improvement in SNR (ISNR): validates the performance of the image
restoration algorithm.
2
f (i, j ) y (i, j )
i j
ISNR 10log10
fˆ (i, j )
2
f (i, j )
i j

where fˆ (i, j ) is the restored image.

Both BSNR and ISNR can only be used for simulation with artificial data.

One dimensional discrete degradation model. Circular convolution

Suppose we have a one-dimensional discrete signal f (i ) of size A

samples f (0), f (1),, f ( A 1) , which is due to a degradation process. The
degradation can be modeled by a one-dimensional discrete impulse response
h(i ) of size B samples. If we assume that the degradation is a causal function

we have the samples h(0), h(1),, h( B 1) .We form the extended versions of
f (i ) and h(i ) , both of size M A B 1 and periodic with period M . These

can be denoted as f e (i ) and he (i ) . For a time invariant degradation process

we obtain the discrete convolution formulation as follows
M 1
ye (i ) f e (m)he (i m) ne (i )
m 0

Using matrix notation we can write the following form

f e ( 0)
f e (1)
y Hf n f ,

f e ( M 1)

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 62

Image Processing 10EC763

he (0) he ( 1)  he ( M 1)
he (1) he (0)  he ( M 2)
H
(M M)    
he ( M 1) he ( M 2)  he (0)

At the moment we decide to ignore the external noise n . Because h is

periodic with period M we have that
he (0) he ( M 1)  he (1)
he (1) he (0)  he ( 2)
H
(M M)    
he ( M 1) he ( M 2)  he (0)
We define (k ) to be
2 2
(k ) he (0) he ( M 1) exp( j k ) he ( M 2) exp( j 2k ) 
M M
2
he (1) exp[ j(M 1)k ], k 0,1, , M 1
M
2 2
Because exp[ j (M i)k ] exp( j ik ) we have that
M M
(k ) MH (k )
H (k ) is the discrete Fourier transform of he (i) .

I define w(k ) to be
1
2
exp( j k)
w( k ) M

2
exp[ j ( M 1)k ]
M
It can be seen that
Hw (k ) (k )w(k )

This implies that (k ) is an eigenvalue of the matrix H and w(k ) is its

corresponding eigenvector.

We form a matrix w whose columns are the eigenvectors of the matrix H ,

that is to say

W w(0) w(1)  w( M 1)

2 1 2
w(k , i ) exp j ki and w 1 (k , i ) exp j ki
M M M
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 63

Image Processing 10EC763

We can then diagonalize the matrix H as follows

H WDW -1 D W-1HW
where
(0) 0
(1)
D

0 ( M 1)

Obviously D is a diagonal matrix and

D(k , k ) (k ) MH (k )

If we go back to the degradation model we can write

y Hf y WDW -1f W -1y DW 1f Y (k ) MH (k ) F (k ), k 0,1, , M 1

Y (k ), H (k ), F (k ), k 0,1, , M 1 are the M sample discrete Fourier

transforms of y(i), h(i), f (i), respectively. So by choosing (k ) and w(k ) as

above and assuming that he (i ) is periodic, we start with a matrix problem and
end up with M scalar problems.

Two dimensional discrete degradation model. Circular convolution

Suppose we have a two-dimensional discrete signal f (i, j ) of size A B

samples which is due to a degradation process. The degradation can now be
modeled by a two dimensional discrete impulse response h(i, j ) of size C D
samples. We form the extended versions of f (i, j ) and h(i, j ) , both of size
M N , where M A C 1 and N B D 1 , and periodic with period
M N . These can be denoted as f e (i, j ) and he (i, j ) . For a space invariant

degradation process we obtain

M 1N 1
ye (i, j ) f e (m, n)he (i m, j n) ne (i, j )
m 0n 0

Using matrix notation we can write the following form

y Hf n
where f and y are MN dimensional column vectors that represent the
lexicographic
ordering of images f (i, j ) and h (i, j ) respectively.
Downloaded by eSudheer Yadav e([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 64

Image Processing 10EC763

H0 HM 1  H1
H1 H0  H2
H
   
HM 1 HM 2  H0

he ( j,0) he ( j, N 1)  he ( j,1)
he ( j,1) he ( j,0)  he ( j,2)
Hj
   
he ( j, N 1) he ( j, N 2)  he ( j,0)

The analysis of the diagonalisation of H is a straightforward extension of the

one-
dimensional case. In that case we end up with the following set of M N
scalar problems.
Y (u, v) MNH (u, v) F (u, v)( N (u, v))

u 0,1,, M 1, v 0,1,, N 1

In the general case we may have two functions f (i), A i B and

h(i), C i D , where A, C can be also negative (in that case the functions are

non-causal). For the periodic convolution we have to extend the functions

from both sides knowing that the convolution is
g (i) h(i) f (i), A C i B D.

Direct deterministic approaches to restoration

Inverse filtering

The objective is to minimize

2 2
J (f ) n(f ) y Hf
We set the first derivative of the cost function equal to zero
J (f )
0 2H T ( y Hf ) 0
f
f ( H T H) -1 H T y
1
If M N and H exists then
f H -1y
According to the previous analysis if H (and therefore H-1 ) is block circulant
the aboveDownloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 65

Image Processing 10EC763

[
problem can be solved as a set of M N scalar problems as follows

H (u, v )Y (u, v ) 1 H (u, v )Y (u, v )

F (u, v ) 2
f (i , j ) 2
H ( u, v ) H ( u, v )

Computational issues concerning inverse filtering

(I) Suppose first that the additive noise n(i, j ) is negligible. A problem arises if
H (u, v ) becomes very small or zero for some point (u, v) or for a whole

region in the (u, v) plane. In that region inverse filtering cannot be applied.
Note that in most real applications H (u, v ) drops off rapidly as a function of
distance from the origin. The solution is that if these points are known they
can be neglected in the computation of F (u, v) .
[

(II) In the presence of external noise we have that

H (u, v) Y (u, v) N (u, v) H (u, v)Y (u, v) H (u, v) N (u, v)
Fˆ (u, v) 2 2 2
H (u, v) H (u, v) H (u, v)
N (u, v)
Fˆ (u, v) F (u, v)
H (u , v)
If H (u, v ) becomes very small, the term N (u, v ) dominates the result. The
solution is again to carry out the restoration process in a limited
neighborhood about the origin where H (u, v ) is not very small. This
procedure is called pseudoinverse filtering. In that case we set
H (u , v)Y (u , v)
2
H (u , v) T
H (u , v)
Fˆ (u , v)
0 H (u , v) T

The threshold T is defined by the user. In general, the noise may very well
possess large components at high frequencies (u, v) , while H (u, v ) and
Y (u, v) normally will be dominated by low frequency components.

Constrained least squares (CLS) restoration

It refers to a very large number of restoration algorithms.The problem can be

formulated as follows.
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 66

Image Processing 10EC763

minimize
2 2
J (f ) n(f ) y Hf
subject to
2
Cf
where Cf is a high pass filtered version of the image. The idea behind the
above constraint is that the highpass version of the image contains a
considerably large amount of noise! Algorithms of the above type can be
handled using optimization techniques. Constrained least squares (CLS)
restoration can be formulated by choosing an f to minimize the Lagrangian
2 2
min y Hf Cf

Typical choice for C is the 2-D Laplacian operator given by

0.00 0.25 0.00

C 0.25 1.00 0.25
0.00 0.25 0.00

represents either a Lagrange multiplier or a fixed parameter known as

regularisation parameter and it controls the relative contribution between
2 2
the term y Hf and the term Cf . The minimization of the above leads to
the following estimate for the original image
1
f HT H CTC HT y

Computational issues concerning the CLS method

Choice of

The problem of the choice of has been attempted in a large number of

studies and different techniques have been proposed. One possible choice is
based on a set theoretic approach: a restored image is approximated by an
image which lies in the intersection of the two ellipsoids defined by
2
Qf |y {f | y Hf E 2 } and
2 2
Qf {f | Cf }
The center of one of the ellipsoids which bounds the intersection of Q f |y and
Qf , is given by the equation
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 67

Image Processing 10EC763

1
f HT H CTC HT y
with (E / ) 2 . Another problem is then the choice of E 2 and 2
. One
choice
[

could be
1
BSNR

Comments

With larger values of , and thus more regularisation, the restored image
tends to have more ringing. With smaller values of , the restored image
tends to have more amplified noise effects. The variance and bias of the error
image in frequency domain are
2
2
M N H ( u, v )
Var ( ) n
u 0v 0 2 2 2
H (u, v ) C ( u, v )
2 2 4
2
M 1N 1 F ( u, v ) C (u, v )
Bias ( ) n
u 0 v 0 2 2 2
H ( u, v ) C ( u, v )

The minimum MSE is encountered close to the intersection of the above

functions. A good choice of is one that gives the best compromise between
the variance and bias of the error image.

Iterative deterministic approaches to restoration

They refer to a large class of methods that have been investigated extensively
over the last decades. They possess the following advantages.
There is no need to explicitly implement the inverse of an operator. The
restoration process is monitored as it progresses. Termination of the
algorithm may take place before convergence.
The effects of noise can be controlled in each iteration.
The algorithms used can be spatially adaptive.
The problem specifications are very flexible with respect to the type of
degradation. Iterative techniques can be applied in cases of spatially varying
or nonlinear degradations or in cases where the type of degradation is
completely unknown (blind restoration).
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 68

Image Processing 10EC763

In general, iterative restoration refers to any technique that attempts to

minimize a function of the form M (f ) using an updating rule for the partially
restored image.

Least squares iteration

In that case we seek for a solution that minimizes the function

2
M (f ) y Hf
A necessary condition for M (f ) to have a minimum is that its gradient with
respect to f is equal to zero. This gradient is given below
M (f )
f M (f ) 2( H T y H T Hf)
f
and by using the steepest descent type of optimization we can formulate an
iterative rule as follows:
f0 HT y
M (f k )
fk 1 fk fk H T (y Hf k ) HTy (I H T H)f k
fk

Constrained least squares iteration

In this method we attempt to solve the problem of constrained restoration

iteratively. As already mentioned the following functional is minimized
2 2
M (f , ) y Hf Cf
The necessary condition for a minimum is that the gradient of M (f , ) is
equal to zero. That gradient is
(f ) f M (f , ) 2[( H T H C T C)f H T y]
The initial estimate and the updating rule for obtaining the restored image are
now given by
f0 HT y
fk 1 fk [H T y ( H T H CT C)fk ]
It can be proved that the above iteration (known as Iterative CLS or
Tikhonov-Miller Method) converges if
2
0
m ax

where m ax is the maximum eigenvalue of the matrix

(H T H C T C)
If the matrices H and C are block-circulant the iteration can be implemented
in the frequency domain.
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 69

Image Processing 10EC763

Projection onto convex sets (POCS)

The set-based approach described previously can be generalized so that any

number of prior constraints can be imposed as long as the constraint sets are
closed convex. If the constraint sets have a non-empty intersection, then a
solution that belongs to the intersection set can be found by the method of
POCS. Any solution in the intersection set is consistent with the a priori
constraints and therefore it is a feasible solution.
Let Q1 , Q2 ,, Qm be closed convex sets in a finite dimensional vector space,
with P1 , P2 ,, Pm their respective projectors. The iterative procedure
fk 1 P1P2 , Pmfk
converges to a vector that belongs to the intersection of the sets
Qi , i 1,2,, m , for any starting vector f0 . An iteration of the form

fk 1 P1P2fk can be applied in the problem described previously, where we

seek for an image which lies in the intersection of the two ellipsoids defined
by
2 2
Qf |y {f | y Hf E 2 } and Qf {f | Cf 2
}
The respective projections P1f and P2f are defined by
1
P1f f λ1 I λ1HTH HT (y Hf )
1
P2f [I λ2 I λ2CTC CTC]f
[
Spatially adaptive iteration

The functional to be minimized takes the form

2 2
M (f , ) y Hf W1 Cf W2

where
2
y Hf W1 ( y Hf) T W1 ( y Hf)
2
Cf W2 (Cf) T W2 (Cf)

W1 ,W2 are diagonal matrices, the choice of which can be justified in various
ways. The entries in both matrices are non-negative values and less than or
equal to unity. In that case
(f ) f M (f , ) (H T W1 H C T W2 C)f H T W1 y
A more specific case is
2 2
M (f , ) y Hf Cf W
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 70

Image Processing 10EC763

where the weighting matrix is incorporated only in the regularization term.

This method is known as weighted regularised image restoration. The
entries in matrix W will be chosen so that the high-pass filter is only
effective in the areas of low activity and a very little smoothing takes place in
the edge areas.

Robust functionals

Robust functionals allow for the efficient supression of a wide variety of

noise processes and permit the reconstruction of sharper edges than their
quadratic counterparts. We are seeking to minimize
M (f , ) Rn ( y Hf ) Rx Cf

Rn (), Rx () are referred to as residual and stabilizing functionals respectively.

Computational issues concerning iterative techniques

(I) Convergence

The contraction mapping theorem usually serves as a basis for establishing

convergence of iterative algorithms. According to it iteration
f0 0
fk 1 fk (f k ) (f k )
converges to a unique fixed point f , that is, a point such that (f ) f , for
any
initial vector, if the operator or transformation (f ) is a contraction. This
means
that for any two vectors f1 and f 2 in the domain of (f ) the following
relation
holds
( f1 ) (f 2 ) f1 f2
with 1 and any norm. The above condition is norm dependent.

(II) Rate of convergence

The termination criterion most frequently used compares the normalized
change in energy at each iteration to a threshold such as
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 71

Image Processing 10EC763

2
fk 1 fk 6
2
10
fk

Stochastic approaches to restoration

Wiener estimator (stochastic regularisation)

The image restoration problem can be viewed as a system identification

problem as follows:

f (i, j ) y(i, j )
W
fˆ (i, j ) H

n(i, j )

The objective is to minimize the following function

E{(f fˆ ) T (f fˆ )}
To do so the following conditions should hold:

(i) E{fˆ} E{f } E{f } WE{y}

(ii) The error must be orthogonal to the observation about the mean

E{(fˆ f )( y E{y}) T } 0
From (i) and (ii) we have that
E{( Wy f )( y E{y}) T } 0 E{( Wy E{f } WE{y} f )( y E{y}) T } 0
E{[ W( y E{y}) (f E{f })]( y E{y}) T } 0
~
If ~
y y E{y} and f f E{f } then
~ T
y f )~
E{( W~ y } 0
~ ~ T
E{W~y~y T } E{f ~
y T } WE{~y~y T } E{f ~
y } WR~y~y R ~f~y

If the original and the degraded image are both zero mean then R ~y~y R yy and
R ~f ~y R fy . In that case we have that WR yy R fy . If we go back to the
degradation model and find the autocorrelation matrix of the degraded image
then we get that

y Hf n yT f THT nT
E{yy T } HR ff HT R nn R yy
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 72

Image Processing 10EC763

E{fy T } Rff HT Rfy

From the above we get the following results

1
W RfyR yy Rff HT (HR ff HT Rnn ) 1

fˆ R ff H T ( HR ff H T R nn ) 1 y

Note that knowledge of R ff and R nn is assumed. In frequency domain

S ff (u, v ) H (u, v )
W ( u, v ) 2
S ff (u, v ) H (u, v ) S nn (u, v )
S ff (u, v ) H (u, v )
Fˆ (u, v ) 2
Y ( u, v )
S ff (u, v ) H (u, v ) S nn (u, v )

Computational issues

The noise variance has to be known, otherwise it is estimated from a flat

region of the observed image. In practical cases where a single copy of the
degraded image is available, it is quite common to use S yy (u, v ) as an

estimate of S ff (u, v ) . This is very often a poor estimate.

Wiener smoothing filter

In the absence of any blur, H (u, v ) 1 and

S ff (u, v ) ( SNR )
W ( u, v )
S ff (u, v ) S nn (u, v ) ( SNR ) 1
(i) ( SNR ) 1 W (u, v) 1
(ii) ( SNR ) 1 W (u, v) ( SNR )
(SNR ) is high in low spatial frequencies and low in high spatial frequencies
so W (u, v ) can be implemented with a lowpass (smoothing) filter.

5.1.3 Relation with inverse filtering

1
If S nn (u, v ) 0 W ( u, v ) which is the inverse filter
H ( u, v )
1
H ( u, v ) 0
H ( u, v )
If S nn (u, v) 0 lim W (u, v )
Snn 0
0 H ( u, v ) 0

which is the pseudoinverse filter.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 73

Image Processing 10EC763

5.1.4 Iterative Wiener filters

They refer to a class of iterative procedures that successively use the Wiener
filtered signal as an improved prototype to update the covariance estimates of
the original image as follows.
Step 0: Initial estimate of R ff
R ff (0) R yy E{yy T }
Step 1: Construct the i th restoration filter
W(i 1) R ff (i )H T ( HR ff (i )H T R nn ) 1

Step 2: Obtain the (i 1) th estimate of the restored image

fˆ (i 1) W(i 1) y
Step 3: Use fˆ (i 1) to compute an improved estimate of R ff given by
R (i 1) E{fˆ (i 1)fˆ T (i 1)}
ff
Step 4: Increase i and repeat steps 1,2,3,4.

Recommended Questions

1. Explain the importance process in image restoration process in image

processing. Explain any four important noise probability density functions.
2. Discuss the importance of adaptive filters in image restoration system.
Highlight the working of adaptive median filters.
3. Explain adaptive median filter and its advantages.
4. How do you reduce the periodic noise using frequency domain filters?
5. Derive the expression for observed image when the degradations are linear
position invariant.
6. With a block diagram, briefly explain the image model of degradation-
restoration process.
7. Explain notch reject filters. How can we obtain the notch filter that pass rather
than suppressing the frequency in notch area?
8. Explain the Weiner filtering method of restoring images.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 74

Image Processing 10EC763

UNIT – 8
Color Fundamentals
The characteristics generally used to distinguish one color from another are
brightness, hue, and saturation
brightness: the achromatic notion of intensity.
hue: dominant wavelength in a mixture of light waves, represents dominant
color as perceived by an observer.
saturation: relative purity or the amount of white light mixed with its hue.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 75

Image Processing 10EC763

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 76

Image Processing 10EC763

Color Models

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 77

Image Processing 10EC763

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 78

Image Processing 10EC763

Pseudo color Image Processing.,

Gray level to Color Transformation

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 79

Image Processing 10EC763

processing basics of full color image processing

Let c represent an arbitrary vector in RGB color space:

cR R
c cG G
cB B
At coordinates (x, y),
cR ( x , y ) R ( x, y )
c( x, y) cG ( x, y) G( x, y)
cB ( x , y ) B ( x, y )

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 80

Image Processing 10EC763

RGB

The RGB colour model relates very closely to the way we perceive colour
with the r, g and b receptors in our retinas. RGB uses additive colour mixing
and is the basic colour model used in television or any other medium that
projects colour with light. It is the basic colour model used in computers and
for web graphics, but it cannot be used for print production.

The secondary colours of RGB – cyan, magenta, and yellow – are formed by
mixing two of the primary colours (red, green or blue) and excluding the
third colour. Red and green combine to make yellow, green and blue to make
cyan, and blue and red form magenta. The combination of red, green, and
blue in full intensity makes white.

In Photoshop using the “screen” mode for the different layers in an image
will make the intensities mix together according to the additive colour mixing
model. This is analogous to stacking slide images on top of each other and
shining light through them.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 81

Image Processing 10EC763

CMYK The 4-colour CMYK model used in printing lays down overlapping
layers of varying percentages of transparent cyan (C), magenta (M) and
yellow (Y) inks. In addition a layer of black (K) ink can be added. The
CMYK model uses the subtractive colour model.

Gamut The range, or gamut, of human colour perception is quite large. The
two colour spaces discussed here span only a fraction of the colours we can
see. Furthermore the two spaces do not have the same gamut, meaning that
converting from one colour space to the other may cause problems for
colours in the outer regions of the gamuts.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 82

Image Processing 10EC763

The HSI color space

The HSI color space is very important and attractive color model for image
processing applications because it represents color s similarly how the human
eye senses colors.

The HSI color model represents every color with three components: hue ( H
), saturation ( S ), intensity ( I ). The below figure illustrates how the HIS
color space represents colors.

The Hue component describes the color itself in the form of an angle between
[0,360] degrees. 0 degree mean red, 120 means green 240 means blue. 60
degrees is yellow, 300 degrees is magenta.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 83

Image Processing 10EC763

The Saturation component signals how much the color is polluted with white
color. The range of the S component is [0,1].

The Intensity range is between [0,1] and 0 means black, 1 means white.

As the above figure shows, hue is more meaningful when saturation

approaches 1 and less meaningful when saturation approaches 0 or when
intensity approaches 0 or 1. Intensity also limits the saturation values.

To formula that converts from RGB to HSI or back is more complicated than
with other color models, therefore we will not elaborate on the detailed
specifics involved in this process.

RGB CMY CMYK

HSI HSV L*a*b

XYZ YIQ YUV

Conversion from RGB to HSI

Given the intensities of the three primaries RGB of a color, we can find its
HSV representation using different models. Here we use the RGB plane of
the cube to find the corresponding HSV. The three vertices are represented

by , and , and the three components of the given color is

represented by a 3D point . We also assume the intensities

are normalized so that the , and values are between 0 and 1, so that
point is inside or on the surface of the color cube.

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 84

Image Processing 10EC763

Determine the intensity I:

One of the definitions of intensity is

Determine the hue H:

First find the intersection of the color vector with

the RGB triangle :

This point is on the RGB triangle as . Here

we assume the point is inside the triangle formed by points , ,

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 85

Image Processing 10EC763

and . The hue is the angle formed by the vectors and .

Consider the dot product of these two vectors:

where , and , and

Now the hue angle can be found to be

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 86

Image Processing 10EC763

If , then .

Determine S:

The saturation of the colors on any of the three edges of the

RGB triangle is defined as 1 (100% saturated), and the

saturation of is zero. Denote as the

intersection of the extension of line with the edge. If the

normalized color is , , and if , .

The saturation of any color point between and is

defined as

Here it is assumed that point is inside the triangle so

that . Yadav ([email protected])

Downloaded by Sudheer
 lOMoARcPSD|49982872

Smartworld.asia 87

Image Processing 10EC763

In general

Or

Conversion from HSI to RGB

Consider three possible cases in terms of the hue angle :

(p inside )

From , we get

Also we can get

and

Given , we can get from . As

we have

Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 88

Image Processing 10EC763

(p inside )

(p inside )

CMYK_CONVERT

The CMYK_CONVERT procedure converts from the CMYK (cyan-

magenta-yellow-black) color model to RGB (red-green-blue) and vice versa.

The procedure uses the following method to convert from CMYK to RGB:

R = (255 - C) (1 - K/255)

G = (255 - M) (1 - K/255)

B = (255 - Y) (1 - K/255)
Downloaded by Sudheer Yadav ([email protected])
 lOMoARcPSD|49982872

Smartworld.asia 89

Image Processing 10EC763

To convert from RGB to CMYK, the procedure uses the following method:

K = minimum of (R, G, B)

C = 255 [1 - R/(255 - K)] (if K=255 then C=0)

M = 255 [1 - G/(255 - K)] (if K=255 then M=0)

Y = 255 [1 - B/(255 - K)] (if K=255 then Y=0)

In both cases the CMYK and RGB values are assumed to be in the range 0 to
255.

Note
There is no single method that is used for CMYK/RGB conversion. The
method used by CMYK_CONVERT is the simplest and, depending on
printing inks and screen colors, might not be optimal in all situations.

This routine is written in the IDL language. Its source code can be found in
the file cmyk_convert.pro in the libsubdirectory of the IDL distribution.

Syntax

CMYK_CONVERT, C, M, Y, K, R, G, B [, /TO_CMYK]

Arguments

C, M, Y, K

To convert from CMYK to RGB, set these arguments to scalars or arrays

containing the CMYK values in the range 0-255. To convert from RGB to
CMYK (with the TO_CMYK keyword set), set these arguments to named
variables that will contain the converted values.

R, G, B Downloaded by Sudheer Yadav ([email protected])

 lOMoARcPSD|49982872

Smartworld.asia 90

Image Processing 10EC763

To convert from CMYK to RGB, set these arguments to named variables that
will contain the converted values. To convert from RGB to CMYK (with
the TO_CMYK keyword set), set these arguments to scalars or arrays
containing the RGB values.

Keywords

TO_CMYK

If this keyword is set, the values contained in the RGB arguments are
converted to CMYK. The default is to convert from CMYK to RGB.

Recommended Questions

1. Explain the colour models.

2. Explain the following order statistics filters, indicating their uses.
i) Median filter ii) max filter iii) min filter.
3. Explain the RGB colour model.
4. Write a note on the following pseudo image processing techniques.
i) Intensity slicing
ii) Graylevel to colour transformations.
5. Write steps involved in converting colours from RGB to HSI and vice versa.
6. Explain pseudocolour image processing in brief.
7. Write short notes on i) weiner filtering ii) Inverse filtering

Downloaded by Sudheer Yadav ([email protected])