Introduction

January 10, 2018

Digital Image Processing
We are talking of a digital image - image captured by a digital
camera not the old fashioned film. Digital Image Processing refers
to the processing of a two dimensional image by a digital computer.
Similar terms: Pattern Recognition, Computer Graphics, Computer
Vision.
DIP: Take any image as input; process this image to improve some
specific features or extract some features of the image. Input is an
image ; after some processing the output is also an image.
Computer Graphics: We generate some pictures like lines, shadings
etc. We specify the descriptions of the object and after processing
these descriptions, the computer generates an [Link] is a set
of descriptions and the output after processing is an image. This is
a synthetic image.
Pattern Recognition: Input is an image and we want to extract
some meaningful description of the features. Example is
Classification of Satellite images.
DIP, Computer Graphics, Pattern Recognition, Computer
Vision

Computer Vision: This must be preceded by DIP. Given any image

it is processed by DIP and we generate some properties and
recognise the different objects in the image. This is one of the
basics of Artificial Intelligence. Robotics. Chandrayaan II, Venus
Explorer.
DIP,Computer Graphics, Pattern Recognition,Computer
Vision

Computer Vision is DIP + Extraction of Information + Decisions

Applicatons of DIP

1. Medical imaging - X-rays, sonography, all kinds of scans; two

images are compared before and after treatment to detect the
changes.
2. Remote Sensing - Satellite (some times aircrafts and balloons)
image the earth; from the images, using DIP techniques lot of
information can be derived;
3. Meteorology - The earth and the surrounding atmosphere is
imaged from a vantage point in space which is outside the
atmosphere. Hence, the movement of the clouds can be clearly
seen and this taken over successive intervals can be used for
making weather prediction;
4. Industrial applications - Cracks and deformations inside the
structure of buildings and bridges can be detected;
Applications of DIP

5. Robotics - classic example is the Moon rover and the rover in

Mars and Venus. They do take the images and send them back to
earth for analysis, but sometimes the analysis is done onboard. In
fact there is a scientific laboratory in the rover in Venus, which
does the spectroscopic analysis of the soil scooped by the robotic
arm and the results are sent back to earth;
6. Arts - Old paintings of the masters like da Vinci, get
deteriorated over a period of time. Such paintings can be restored
by techniques called InPainting etc.
7. Sports - The DRS system used in Cricket - more than one
camera images the trajectory of the ball and then mathematical
modeling is used to predict the future course of the ball to see if it
would have hit the stumps if allowed to proceed further.
Why do we process an image?

The subject has developed in respoonse to three major problems

concerned with pictures:
1. picture digitization and coding to facilitate transmission
printing and storage of pictures- first needed when pictures
were transmitted from London to US for newspaper
publication
2. picture enhancement and restoration - the need was acutely
felt in attempts to interpret the images sent by space probes -
try to see the first pictures sent by Armstrong from the moon
3. picture segmentation and description - used for medical
applications - stepping stones for computer vision
Basic definitions

What is an image?
A panchromatic image is a 2D light intensity function, f (x, y),
where x, y are spatial coordinates and the value of f at (x, y) is
proportional to the brightness of the scene at that point.
Panchromatic means single band image.
If we have a multispectral image - means in general a color image -
f (x, y) is a vector, each component of which indicates the
brightness of the scene at the point (x, y) in the corresponding
spectral band.
Basic definitions

What is a spectral band?

A color band is a range of wavelengths of the electromagnetic
spectrum over which the sensors we use to capture an image have
nonzero sensitivity. Typical color iamges consist of three color
bands - primary colors are Red(R),Green(G) and Blue(B). This
means that the image has been captured by three sets of sensors
each set made to have a different sensitivity function.
Some Definitions
What is a Digital Image?
We have seen that an image is a function of two variables. A digital
image is an image f (x, y) that has been discretized both in spatial
coordinates and brighness. It is represented by a 2D integer array,
or a series of 2D arrays one for each band. The digitized brightness
value is called a grey level or grey value or grey count.
Each element of the array, which represents a spatial location is
called a pixel or pel. Thus a digital image looks like

 
f (1, 1) f (1, 2) ... f (1, N )
 f (2, 1) f (2, 2) ... f (2, N ) 
f (x, y) =   , 0 ≤ f (x, y) ≤ G
 ... ... ... ... 
f (N, 1) f (N, 2) ... f (N, N )

where normally G and N are powers of 2.

Of course the number of rows and columns of the matrix need not
be same.
Some basic concepts
How is a digital image formed?
Each pixel of an image corresponds to some part of a physical
object in the 3D world. This object is illuminated by some light,
which is partly absorbed and partly reflected. Part of the reflected
light reaches the sensor used to image the scene and is responsible
for the values recorded by these sensors. One of these sensors
makes up a pixel and its field of view corresponds to a small patch
in the imaged sene. The recorded value by each sensor depends on
its sentitivy curve. When a photon of a certain wavelength reaches
the sensor, its energy is multiplied with the value of the sensitivity
curve of the sensor at that wavelength and is accumulated. The
total energy collected by the sensor during the exposure time is
eventually used to compute the grey value of the pixel that
corresponds to this sensor.
Note that the sensor records the amount of energy falling on it and
hence theoretically varies from 0 to ∞. When no energy falls, the
patch of physical world is absolute black.
Some basic concepts

Thus the brightness values of different pixels have been crated by

using the energies recorded by the corresponding sensors. They
have significance only relative to each other and they are
meaningless in absolute terms. So, pixel values between different
images should only be compared if either care has been taken for
the physical processes used to form the two images to be identical,
or the brighness values of the two images have somehow been
normalised so that the effects of the different physical processes
have been removed. In that case, we say that the sensors are
calibrated.
Some Digital Image

Figure: Processed Visible and IR images of 3D

Some Digital Images

Figure: X-ray; Boat; Cameraman; Lena

Digital Image

What is a digital image?

An image is a two dimensional representation. We see variations in
the picture because of difference in the amount of light reflected.
(Refer to Optics as to how we see anything). The amount of light
reflected is called Intensity of image. If at a particular location, it
is very bright, then the intensity is very high, and if it is dark, then
intensity is low. Similar to what we say in everyday language.
Intensity is also called gray values. So a two dimensional image can
be thought of as a rectangular array (rectangular if height and
width are different)of elements, with each element having a
definite intensity or gray value. These elements are called Picture
Element or pixel for short.
Digital Images
Digital Images

Every image can be thought of as a two dimensional matrix of

numbers, which represent the intensity at that particular position.
The size of the matrix determines the spatial accuracy you can get
out of the image. This is in fact called the Resolution. So you
have a scene and you divide into small elements, each element
representing the average of the intensity of the small area under
consideration. Thus the gray value at any point does not represent
the intensity at one point but the average over a small elementary
area.
What about intensity, and what are the numbers in the box on the
right side of the previous image. They represent the gray count at
those pixel locations.
Digital Images

Original and Condensed images

Digital Images

Thus for getting digital images, we must sample the data in spatial
directions. Thus we generate a two dimensional grid ( and hence a
matrix). At these locations the intensity values are assigned.
Continuous data being put at discrete locatons is called Sampling.
How close should the samples be? This introduces the concept of
Resolution. Closer the samples, better is the visual impact.
Effect of quantisation

8bits 7bits 6bits 5bits

4bits 3bits 2bits 1bits
boat image
Compressed and zoomed images
ImageJ usage