Digital Image Processing

Assignment- Week 1
TYPE OF QUESTION: MCQ/MSQ
Number of questions: 10 Total mark: 10 X 2 = 20
______________________________________________________________________________

QUESTION 1:
In frequency domain, what is the equivalent operation of product of two functions in spatial
domain?
a) correlation
b) convolution
c) Fourier transform
d) fast Fourier transform
Correct Answer: b
Detailed Solution:

Correlation in frequency domain is the equivalent operation of product of two functions in spatial
domain.

____________________________________________________________________________

QUESTION 2:

An image depicts IIT Kharagpur’s main building on a rainy day of cyclone “Remal”. Removal of
rain streaks from the image is an example of__________

a) Image compression
b) Image restoration
c) Image enhancement
d) Image Analysis

Correct Answer: c. Image enhancement

Detailed Solution:

The process of removal of rain streaks, fog or hailstorm for better visibility is an example of image
enhancement. Note the it is not image restoration since original image did not undergo corruption
or damage due to time or other physical elements (exposure of film negatives to sunlight or picture
getting torn and patched up by glue). Hence option c is correct.
______________________________________________________________________________
QUESTION 3:
A photograph, W inches in width and H inches in height was scanned at 1000 dpi. The scanned
image has a resolution of 100 pixels in width and 200 pixels in height What is the physical size of
the image.
a. W=0.1 inches, H=0.2 inches
b. W=0.2 inches, H=0.1 inches
c. W=0.1 inches, H=0.1 inches
d. Some information is missing

Correct Answer: a.

Detailed Solution:

Image scanned at 1000 dpi.

Height and width= 200 pixels and 100 pixels.
Image size= (200/1000) X (100/1000) = 0.2 inches X 0.1 inches.
Hence option a is correct
______________________________________________________________________________

QUESTION 4:
Area under unit impulse function,

∞, 𝑡𝑡 = 0
𝛿𝛿(𝑡𝑡) = � .
0, 𝑡𝑡 ≠ 0

is _______.

a) 1
b) undefined
c) infinity
d) 0

Correct Answer: a.

Detailed Solution:

∞, 𝑡𝑡 = 0
From the definition of unit impulse, 𝛿𝛿(𝑡𝑡) = � .
0, 𝑡𝑡 ≠ 0

Hence option a is correct.

______________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 5:

What is the number of bits required to store a 100 × 100 grayscale image with 256 gray levels?
a) 2560000
b) 10000
c) 80000
d) 240000

Correct Answer: (c)

Detailed Solution:

Number of gray levels=256=28. So we would require 8 bits per pixel.

Image size = 100 × 100 = 10,000 pixels. Number of channels = 1.
Bytes required = No. of pixels × bytes per pixel × No. of channels
= 10,000 × 8 × 1
= 80,000 bits.
____________________________________________________________________________

QUESTION 6:
+∞
What will be the value of delta function, ∫−∞ 𝑓𝑓(𝑡𝑡)𝛿𝛿(𝑡𝑡 − 2)𝑑𝑑𝑑𝑑 ?

a. Infinity
b. 𝑓𝑓(2)
c. 𝑓𝑓(𝑡𝑡 − 2)
d. 𝑓𝑓(0)

Correct Answer: b
Detailed Solution:

∞, 𝑡𝑡 = 0
𝛿𝛿(𝑡𝑡) = � . Using the time shifting property,
0, 𝑡𝑡 ≠ 0

∞, 𝑡𝑡 = 2
𝛿𝛿(𝑡𝑡 − 2) = �
0, 𝑡𝑡 ≠ 0
+∞
Therefore ∫−∞ 𝑓𝑓(𝑡𝑡)𝛿𝛿(𝑡𝑡 − 2)𝑑𝑑𝑑𝑑 samples value of 𝑓𝑓(𝑡𝑡) at t=2; 𝑓𝑓(2)

______________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 7:
What is the purpose of using a low pass filter in image digitization?

a. Resizing
b. Sharpening
c. Sampled Signal recovery
d. None of the above

Correct Answer: c
Detailed Solution:

Low pass filter are used for image recovery from sampled signal

______________________________________________________________________________

QUESTION 8:
Choose the appropriate option depending on statements below:
Statement 1: Unit impulse function samples signal at multiple points
Statement 2: Comb function samples signal at a single point

a) Statement 1 is True, Statement 2 is True.

b) Statement 1 is False, Statement 2 is True.
c) Statement 1 is True, Statement 2 is False.
d) Statement 1 is False, Statement 2 is False.

Correct Answer: d
Detailed Solution:

Unit impulse function samples signal at single point, whereas Comb function samples signal at a
multiple points, see your notes
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 9:
Which of the following is not studied under Digital Image Processing?

a) Designing lens geometry and choosing lens material for least chromatic and
spatial aberrations
b) Applying spatial filters with fixed coefficients for Image enhancement
c) Transforming image based on its statistical properties
d) Applying frequency transforms to image for Image compression
Correct Answer: a
Detailed Solution:

The field of optimal lens geometry and lens material is generally studied under optical physics
and material sciences.

______________________________________________________________________________

QUESTION 10:
Choose the correct sequence of operations in Image Digitization

a. Quantization, Sampling, Storage

b. Storage, Sampling, Quantization
c. Sampling, Quantization, Storage
d. None of Above
Correct Answer: c
Detailed Solution:

Refer to Notes

______________________________________________________________________________

************END*******
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Digital Image Processing

Assignment- Week 2
TYPE OF QUESTION: MCQ/MSQ
Number of questions: 10 Total mark: 10 X 2 = 20
______________________________________________________________________________

QUESTION 1:
Which of the following option is true?
Q

a) Q ∈ N4 (P)
b) Q ∈ N8 (P)
c) Q ∈ ND (P)
d) {Q} ⊄ {ND (P)} ∪ {N4 (P)} ∪ {ND (P)}

Correct Answer: d
Detailed Solution:

From the definition of 8-connectedness and diagonal connectedness, The relations, Q ∈ N4 (P)
, Q ∈ N8 (P)and Q ∈ ND (P) do not hold.
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 2:
5 5 5 5 5
5 5 5 5 5
 
Find the 2D convolution of the given matrices using zero padding. Input =  5 5 5 5 5 
 
10 10 10 10 10
10 10 10 10 10
− 1 − 2 − 1
and kernel =  0 0 0  .
 1 2 1 

− 15 − 20 − 20 − 20 − 15
 0
 0 0 0 0 
a) − 15 − 20 − 20 − 20 − 15
 
− 15 − 20 − 20 − 20 − 15
 30 40 40 40 30 

 15 20 20 20 15 
 0
 0 0 0 0 
b) − 15 − 20 − 20 − 20 − 15
 
− 15 − 20 − 20 − 20 − 15
 30 40 40 40 30 

 − 15 − 20 − 20 − 20 − 15 
 0
 0 0 0 0 
c)  15 20 20 20 15 
 
 15 20 20 20 15 
− 30 − 40 − 40 − 40 − 30

 − 15 − 20 − 20 − 20 − 15 
 0
 0 0 0 0 
d)  − 15 − 20 − 20 − 20 − 15 
 
 − 15 − 20 − 20 − 20 − 15 
− 30 − 40 − 40 − 40 − 30
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

− 15 − 20 − 20 − 20 − 15
 0
 0 0 0 0 
Correct Answer: a. − 15 − 20 − 20 − 20 − 15
 
− 15 − 20 − 20 − 20 − 15
 30 40 40 40 30 

Detailed Solution:

For convolution, we have to do either time reversal of input image or the kernel. Let’s perform time
reversal of kernel and move the kernel on the input image.

5 5 5 5 5
5 5 5 5 5
  1 2 1
Input image, I =  5 5 5 5 5  and time reversed kernel= 0
 0 0 
 
10 10 10 10 10 − 1 − 2 − 1
10 10 10 10 10

I(0,0) = 0x1 + 0x2 + 0x1 + 0x0 + 5x0 + 5x0 + 0x(-1) + 5x(-2) + 5x(-1) = -15

I(0,1) = 0x1 + 0x2 + 0x1 + 5x0 + 5x0 + 5x0 + 5x(-1) + 5x(-2) + 5x(-1) = -20

In this we move the time reversed kernel over the image I and finally obtain the convolved image
as
− 15 − 20 − 20 − 20 − 15
 0
 0 0 0 0 
− 15 − 20 − 20 − 20 − 15
 
− 15 − 20 − 20 − 20 − 15
 30 40 40 40 30 

Hence option a is correct.

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 3:
Compute the Euclidean Distance (D1), City-block Distance (D2) and Chessboard distance (D3) for
points p and q, where p and q be (1, 2, 3) and (1, 5, 7) respectively. Give answer in the form (D1,
D2, D3).
a) (5, 3, 4)
b) (5, 4, 3)
c) (7, 5, 4)
d) (5, 7, 4)

Correct Answer: d. (5, 7, 4)

Detailed Solution:

Euclidean distance (D1) =�(𝑥𝑥 − 𝑢𝑢)2 + (𝑦𝑦 − 𝑣𝑣)2 + (𝑧𝑧 − 𝑤𝑤)2 =5

City block distance (D2) =|x-u|+|y-v| +|z-w|=7
Chessboard distance (D3) =max (|x-s|, |y-t|, |z-w|) = 4

Hence option d is correct.

______________________________________________________________________________

QUESTION 4:

What is the minimum bandwidth required to transmit a grayscale image scanned at 100 PPI of size
3 𝑖𝑖𝑖𝑖𝑖𝑖ℎ𝑒𝑒𝑒𝑒 × 4 𝑖𝑖𝑖𝑖𝑖𝑖ℎ𝑒𝑒𝑒𝑒, with a data compression ratio of 8:1, over a communication channel in 0.5
seconds, given that each pixel is digitized using 8 bits?
Hint: Data compression ratio is defined as the ratio between the uncompressed size and compressed
size: Thus, a representation that compresses a file's storage size from 10 MB to 2 MB has a
compression ratio of 10/2 = 5/1
a) 3.84 Gbps
b) 1.92 Gbps
c) 1.48 Gbps
d) 5.12 Gbps

Correct Answer: a

Detailed Solution:
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

First we need to convert continuous signal into discrete signal by the process of sampling and then
digitize using digital bits.
Therefore, size of pixel at a rate of [(3 × 400) × (4 × 400)] × 8 𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 = 15.36 𝐺𝐺𝐺𝐺
Size of compressed data = Uncompressed data/Compression ratio = 15.36/8 = 1.92 𝐺𝐺𝐺𝐺
Bandwidth of Channel = 1.92/0.5 = 3.84 𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺
____________________________________________________________________________

QUESTION 5:
Which of the following is always True?
a) Chessboard distance > City block distance > Euclidean
b) Chessboard distance < City block distance < Euclidean
c) Chessboard distance < Euclidean < City block distance
d) 𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁 𝑜𝑜𝑜𝑜 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎

Correct Answer: d
Detailed Solution:

Given points (𝑥𝑥, 𝑦𝑦) and (𝑢𝑢, 𝑣𝑣)

City block distance (D1) =|x-u|+|y-v|

Euclidean distance (D2) =�(𝑥𝑥 − 𝑢𝑢)2 + (𝑦𝑦 − 𝑣𝑣)2

Chessboard distance (D3) =max (|x-u|, |y-v|)
D2 ≥ D3; (D2=D3 when 𝑦𝑦 = 𝑣𝑣 𝑜𝑜𝑜𝑜 𝑥𝑥 = 𝑢𝑢 𝑜𝑜𝑜𝑜 𝑏𝑏𝑏𝑏𝑏𝑏ℎ)
D1 ≥ D2; (By triangle inequality, D2 is hypotenuse to right angles with sides |x-u| and |y-v| )
Thus D1 ≥ D2≥ D3, Option C is special case which is not always True.
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 6:
Find the value of logical operation XOR for the binary images A and B (A XOR B). Assume 1 to
be foreground and 0 to be background pixels.

Image-A Image-B

Correct Answer: A.

Detailed Solution:

Following the truth table for XOR operation

XYZ
-------
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

001
010
100
111
-------

XOR between and is

Hence option a is correct.

______________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 7:
Perform only first pass of connected component labelling algorithm considering 4-connected
component labelling in following image and choose the correct option. (labelling starts from label
index=1)

1 1 1

1 1

1 1 1

a 1

4 4 4

4 4

4 4 4

3 3 3

3 3

3 3 3
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

4 5 6

7 8

9 10 11

d None of These

Correct Answer: a

Detailed Solution:

Run Connected Component Algorithm as discussed in class.

______________________________________________________________________________

QUESTION 8:

Consider a 2D point [1,1]T in a continuous image. The point is rotated by 30 o in anticlockwise

direction and scaled by 2 both in x-direction and y-direction. The transformed point is given by
a) [(√3 − 1), (√3 + 1)]T
b) [(√3 + 1), (√3 + 1)]T
c) [2√2, 0]T
d) [(√3 − 1), (√3 − 1)]T

Correct Answer: a
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Detailed Solution:

Rotation matrix in homogenous coordinate system for anti-clockwise direction is given as

cosθ -sinθ 0
𝑅𝑅(𝜃𝜃) = � sinθ cosθ 0�.
0 0 1
Scaling matrix α in x direction and β in y-direction is given by
α 0 0
𝑆𝑆 = � 0 β 0�.
0 0 1

𝑥𝑥′ 2 0 0 cos300 -sin300 0 1

�𝑦𝑦′� = �0 2 0� � sin300 cos300 0� �1�.
1 0 0 1 0 0 1 1

On solving this equation, we get option 𝑎𝑎

______________________________________________________________________________

QUESTION 9:
Consider the following transformation matrix T:
αcosθ αsinθ 0
𝑇𝑇 = �−βsinθ βcosθ 0�
0 0 1

a) T rotates point clockwise by angle θ about x axis, scales x-axis by α and y-axis by β
b) T rotates point anticlockwise by angle θ about x axis, scales x-axis by α and y-axis by
β
c) T rotates point clockwise by angle θ about x axis, scales x-axis by β and y-axis by α
d) T rotates point anticlockwise by angle θ about x axis, scales x-axis by β and y-axis
by α

Correct Answer: a
Detailed Solution:

𝑇𝑇 = 𝑆𝑆 𝑅𝑅(θ)

𝑅𝑅(θ) = Rotation Matrix 𝑖𝑖𝑖𝑖 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

For clockwise rotation angle

= −θ

𝑆𝑆 = Scaling Matrix

α 0 0 cos(-θ) -sin(-θ) 0
𝑇𝑇 = � 0 β 0� � sin(-θ) cos(-θ) 0�
0 0 1 0 0 1

α 0 0 cosθ sinθ 0
𝑇𝑇 = � 0 β 0� �-sinθ cosθ 0�
0 0 1 0 0 1

______________________________________________________________________________

QUESTION 10:
An image sensor takes values between 1.0 and 11.0. If it is quantized by a uniform quantizer with
128 levels, what will be the transition (𝑡𝑡𝑘𝑘 ) and reconstruction level (𝑟𝑟𝑘𝑘 )?
10(𝑘𝑘+1) 5
a) 𝑡𝑡𝑘𝑘 = 128
, 𝑘𝑘 = 1,2,3, . . . . ,129; 𝑟𝑟𝑘𝑘 = 𝑡𝑡𝑘𝑘 + 128 , 𝑘𝑘 = 1,2,3, . . . . ,128

10(𝑘𝑘+1) 10
b) 𝑡𝑡𝑘𝑘 = 128
, 𝑘𝑘 = 1,2,3, . . . . ,129; 𝑟𝑟𝑘𝑘 = 𝑡𝑡𝑘𝑘 + 128 , 𝑘𝑘 = 1,2,3, . . . . ,128

10(𝑘𝑘−1) 5
c) 𝑡𝑡𝑘𝑘 = 128
, 𝑘𝑘 = 1,2,3, . . . . ,129; 𝑟𝑟𝑘𝑘 = 𝑡𝑡𝑘𝑘 + 128 , 𝑘𝑘 = 1,2,3, . . . . ,128

10(𝑘𝑘) 5
d) 𝑡𝑡𝑘𝑘 = 128
, 𝑘𝑘 = 1,2,3, . . . . ,129; 𝑟𝑟𝑘𝑘 = 𝑡𝑡𝑘𝑘 + 128 , 𝑘𝑘 = 1,2,3, . . . . ,128

Correct Answer: c
Detailed Solution:

Using the uniform quantizer design, transition and reconstruction levels can be defined as
10(𝑘𝑘−1) 5
𝑡𝑡𝑘𝑘 = 128 , 𝑘𝑘 = 1,2,3, . . . . ,129; 𝑟𝑟𝑘𝑘 = 𝑡𝑡𝑘𝑘 + 128 , 𝑘𝑘 = 1,2,3, . . . . ,128.
______________________________________________________________________________

______________________________________________________________________________

************END*******
 Digital Image Processing
Assignment- Week 3
TYPE OF QUESTION: MCQ/MSQ
Number of questions: 10 Total mark: 10 X 1 = 10
______________________________________________________________________________

QUESTION 1:

The following figure shows an Original Image (Before transformation) in dotted lines and faded
colors and a transformed image in solid lines and dark colors. Choose the correct transformation
matrix

−1 0 0
a) � 0 −1 0�
0 0 1

1 0 0
b) �0 −1 0�
0 0 1

−1 0 0
c) � 0 1 0�
0 0 1

1 0 −1
d) �0 1 −1�
0 0 1

Correct Answer: a
Detailed Solution:

Let x’, y’ be transformed coordinates and x,y be original coordinates.Then,

𝑚𝑚𝑚𝑚′ 𝑎𝑎 𝑏𝑏 𝑐𝑐 𝑥𝑥
�𝑚𝑚𝑚𝑚′� = �𝑑𝑑 𝑒𝑒 𝑓𝑓 � �𝑦𝑦�
𝑚𝑚 𝑔𝑔 ℎ 𝑖𝑖 1
Pixel location in Original (x,y) -> Pixel location in transformed Image (x’,y’)
(0,0) -> (0,0)
(1,0) -> (-1,0)
(0,1) -> (0, -1)
(1,1) -> (-1,-1)
Using above mappings and solving the equations yields
𝑚𝑚𝑚𝑚′ −𝑚𝑚 0 0 𝑥𝑥
�𝑚𝑚𝑚𝑚′� = � 0 −𝑚𝑚 0 � �𝑦𝑦�
𝑚𝑚 0 0 𝑚𝑚 1

Since m is free to assume any value, m=1 yield option a.

____________________________________________________________________________

QUESTION 2:
Given 2 points a ,b in non-standard homogeneous coordinate system,
1 3
1 3
𝑎𝑎 = � � , 𝑏𝑏 = � �
1 3
1 3
select the correct statement
a) Point ‘a’ and ‘b’ gets projected to same point on image plane
b) A unique ray passes through point ‘a’, point ‘b’, center of projection (i.e. pinhole)
and intersects the image plane
c) Both a) and b)
d) Point ‘a’ and ‘b’ get projected to different points on image plane

Correct Answer: c
Detailed Solution:

1 3
1 3
𝑎𝑎 = � � , 𝑏𝑏 = � �
1 3
1 3
When written in standard homogeneous system coordinate system are

1 3/3 1
1 3/3 1
𝑎𝑎 = � � , 𝑏𝑏 = � �=� �
1 3/3 1
1 3/3 1
Thus both the point ‘a’ and ‘b’ are equivalent in standard homogeneous system coordinate system,
therefore there is a unique ray that passes through point ‘a’ and ‘b’ and intersects camera center,
projecting point ‘a’ and ‘b’ at same location on image plane.
______________________________________________________________________________

QUESTION 3:
How many cameras are required to find 3-D point using image coordinates of camera?
a) Atleast 1
b) Atleast 2
c) Atleast 3
d) Atleast 4

Correct Answer: b.
Detailed Solution:

A single camera provide us a ray in 3-D. So in order to find the corresponding 3-D point, we
require at least two cameras such that the intersection of these two rays (one from each camera)
gives us the 3-D point.

Hence option b is correct.

______________________________________________________________________________

QUESTION 4:
For a stereo pair, one camera is shifted along x-axis with respect to other. Imaging plane is XY
plane. If B is the baseline between the cameras, λ is the focal length, and ( x 2 − x1 ) is the disparity
along x-axis then which of the following is true?

λB
a) Z = B +
( x 2 − x1 )
λB
b) Y = λ −
( x 2 − x1 )
λB
c) Z = λ −
( x 2 − x1 )
λB
d) X = λ −
( x 2 − x1 )

Correct Answer: c.
Detailed Solution:

By using standard stereo model, we get

λB
Z= λ−
( x 2 − x1 )

Hence option d is correct.

______________________________________________________________________________

QUESTION 5:
Two identical cameras having focal length of 0.04m are used for stereo imaging. If the camera
displacement along X axis is 6.0 cm, left image point corresponding to a world point W is (0.2 mm,
0.4 mm) and the corresponding right image point is (0.4 mm, 0.4 mm), find out the 3-D location of
W with respect to a world coordinate system aligned with the coordinate system of the left camera.

a) (-60cm, -90cm, 1240 cm)

b) (6cm, 12cm, -1196cm)
c) (6cm, 9cm, 1204cm)
d) (-6cm, -9cm, 120.4cm)
Correct Answer: b
Detailed Solution:

λ =0.04m, x2 = 0.4mm, x1 = 0.2mm, B = 6.0cm. Displacement is along x-axis, so by using standard

expression for depth in stereo geometry,
λB
Z = λ − (𝑥𝑥 , substituting the above values we get Z = -1196cm
2 −𝑥𝑥1 )

For X and Y with respect to left camera, we use the following relations

𝑥𝑥0 𝑦𝑦0
𝑋𝑋 = (𝜆𝜆 − 𝑍𝑍) and 𝑌𝑌 = (𝜆𝜆 − 𝑍𝑍),
𝜆𝜆 𝜆𝜆

Substituting Z = -1196cm we get X = 6cm and Y = 12cm.

______________________________________________________________________________

QUESTION 6:

Which of the following deals with the conversion of three-dimensional world co-ordinate to two
dimensional image?
a. Perspective Transfomation
b. Rotation
c. Linear Transformation
d. Non-linear mapping
Correct Answer: a

Detailed Solution:

Perspective Transfomation deals with the conversion of three dimensional world co-ordinate to two
dimensional image.
______________________________________________________________________________

QUESTION 7:
Which of the following is correct?

I. Perspective transformation is one to one mapping.

II. Perspective transformation is many to one mapping.
III. Geometrical transformations are commutative.

a) Only I
b) Only II
c) I and III
d) II and III

Correct Answer: b.

Detailed Solution:

Perspective transformation mapping of point a 3D world to point c in the image plane is not a one
to one mapping. Rather it is a many to one mapping because all the 3-D points that lies on the
straight line passing through imaging coordinate and camera center is imaged on same imaging
coordinates.

Geometrical transformations are matrix operations, product of two matrices is not always
commutative.

Hence option b is correct.

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

______________________________________________________________________________

QUESTION 8:
For a camera with focal length of 0.10, find out the locus of the points which will be imaged at
location (0.25, 0.5) on the Image plane. Assume the camera coordinate system and world
coordinate system to be perfectly aligned.

a) X = -0.5 + 5Z, Y = -0.25 + 2.5Z

b) X = 0.5 − 5Z, Y = 0.25 - 2.5Z
c) X = 0.25 - 2.5Z, Y = 0.5 − 5Z
d) X = 0.5 - 10Z, Y = 0.5 - 10Z
Correct Answer: c.
Detailed Solution:

By using pin-hole camera geometry equations,

𝑥𝑥0 𝑦𝑦0
𝑋𝑋 = (𝜆𝜆 − 𝑍𝑍) and 𝑌𝑌 = (𝜆𝜆 − 𝑍𝑍), where (𝑋𝑋, 𝑌𝑌, 𝑍𝑍)are world coordinates,
𝜆𝜆 𝜆𝜆

𝜆𝜆is focal length and (𝑥𝑥𝑜𝑜 , 𝑦𝑦𝑜𝑜 ) is image coordinates. Using these two equations we get,
X = 0.25 - 2.5Z, Y = 0.5 − 5Z

______________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 9:
The following figure shows an Original Image (Before transformation) in dotted lines and faded
colors and a transformed image in solid lines and dark colors. Choose the correct transformation
matrix

cos 𝜙𝜙 −sin 𝜙𝜙 0
a) � sin 𝜙𝜙 cos 𝜙𝜙 0�
0 0 1

1 tan 𝜙𝜙 0
b) �0 1 0�
0 0 1

1 0 0
c) �tan 𝜙𝜙 1 0�
0 0 1

cos 𝜙𝜙 sin 𝜙𝜙 0
d) �−sin 𝜙𝜙 cos 𝜙𝜙 0�
0 0 1

Correct Answer: b
Detailed Solution:

Let x’, y’ be transformed coordinates and x,y be original coordinates. Then,

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

𝑚𝑚𝑚𝑚′ 𝑎𝑎 𝑏𝑏 𝑐𝑐 𝑥𝑥
�𝑚𝑚𝑚𝑚′� = �𝑑𝑑 𝑒𝑒 𝑓𝑓 � �𝑦𝑦�
𝑚𝑚 𝑔𝑔 ℎ 𝑖𝑖 1
Pixel location in Original (x,y) -> Pixel location in transformed Image (x’,y’)
(0,0) -> (0,0)
(1,0) -> (1,0)
(0,1) -> (tan 𝜙𝜙, 1)
(1,1) -> (1+tan 𝜙𝜙,1)
Using above mappings and solving the equations yields
𝑚𝑚𝑚𝑚′ 𝑚𝑚 mtan 𝜙𝜙 0 𝑥𝑥
�𝑚𝑚𝑚𝑚′� = � 0 𝑚𝑚 0 � �𝑦𝑦�
𝑚𝑚 0 0 𝑚𝑚 1

Since m is free to assume any value, m=1 yields option b.

______________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 10:
The following figure shows an Original Image (Before transformation) in dotted lines and faded
colors and a transformed image in solid lines and dark colors. Choose the correct transformation
matrix

cos 𝛹𝛹 −sin 𝛹𝛹 0
a) � sin 𝛹𝛹 cos 𝛹𝛹 0�
0 0 1

1 tan 𝛹𝛹 0
b) �0 1 0�
0 0 1

1 0 0
c) �tan 𝛹𝛹 1 0�
0 0 1

cos 𝛹𝛹 sin 𝛹𝛹 0
d) �−sin 𝛹𝛹 cos 𝛹𝛹 0�
0 0 1

Correct Answer: c
Detailed Solution:
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Let x’, y’ be transformed coordinates and x,y be original coordinates. Then,

𝑚𝑚𝑚𝑚′ 𝑎𝑎 𝑏𝑏 𝑐𝑐 𝑥𝑥
�𝑚𝑚𝑚𝑚′� = �𝑑𝑑 𝑒𝑒 𝑓𝑓 � �𝑦𝑦�
𝑚𝑚 𝑔𝑔 ℎ 𝑖𝑖 1
Pixel location in Original (x,y) -> Pixel location in transformed Image (x’,y’)
(0,0) -> (0,0)
(1,0) -> (1, tan 𝛹𝛹)
(0,1) -> (0,1)
(1,1) -> (1, 1+tan 𝛹𝛹)
Using above mappings and solving the equations yields
𝑚𝑚𝑚𝑚′ 𝑚𝑚 0 0 𝑥𝑥
�𝑚𝑚𝑚𝑚′� = �mtan 𝛹𝛹 𝑚𝑚 0 � �𝑦𝑦�
𝑚𝑚 0 0 𝑚𝑚 1

Since m is free to assume any value, m=1 yields option c.

______________________________________________________________________________

______________________________________________________________________________

************END*******
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Digital Image Processing

Assignment- Week 4
TYPE OF QUESTION: MCQ/MSQ
Number of questions: 10 Total mark: 10 X 1 = 10
______________________________________________________________________________

QUESTION 1:

A 9 x 9 image is as given below. Using a modified linear interpolation, find out the intensity at
location (3.3, 3.6).

Y 0 1 2 3 4 5 6 7 8
X
0 20 22 15 18 25 32 16 18 20
1 22 25 28 25 35 32 20 17 15
2 45 35 16 23 44 11 18 24 18
3 35 32 26 46 24 13 25 24 25
4 32 45 23 34 45 43 42 24 16
5 34 45 42 36 27 34 25 27 19
6 43 46 47 34 13 13 24 17 20
7 43 54 34 22 21 35 27 43 29
8 32 43 26 25 27 36 23 56 34

a) 32.42
b) 37.60
c) 35.14
d) None of the above

Correct Answer: c.

Detailed Solution:

Since we are doing interpolation in 2 dimension, we can convert this to one dimension by first
interpolation along the columns and then along a row.

As we are using linear interpolation, so the region of support would be two pixels (left-right or
top-bottom).
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Here we first interpolate along columns to find the values of A and B. Using linear interpolation
equation,

t , 0 ≤ t <1

B0, 2 (t ) =  2 − t , 1 ≤ t < 2 , now for A we shift the peak at 3.3
0, otherwise


 t − 2.3, 2.3 ≤ t < 3.3


B2.3, 2 (t ) = 2 − (t − 2.3), 3.3 ≤ t < 4.3
 0, otherwise


Substituting the values at t=3 and t= 4 (values are known at these locations) in proper
intervals, we get the coefficient (contribution) that has to be multiplied with intensity
values at location
t = 3 and t=4.

0.7, 2.3 ≤ t < 3.3


B2.3, 2 (t ) = 0.3, 3.3 ≤ t < 4.3
 0, otherwise


Now f (3.3,3) = A = ∑ pi Bi − 2.3, 2 (t ) = 0.7 * 46 + 0.3 * 34 = 42.4

Similarly, we calculate f (3.3,4) = B = ∑ pi Bi − 2.3, 2 (t ) = 0.7 * 24 + 0.3 * 45 = 30.3

Now we interpolate along row to get the value of X using A and B. By following similar
procedure of one dimensional linear interpolation,
f (3.3,4.6) = X = ∑ pi Bi − 2.6, 2 (t ) = 0.4 * 42.4 + 0.6 * 30.3 = 35.14

____________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 2:
For the one-dimensional function f(t), given below. Using modified cubic interpolation, find out
the value at location t = 6.3.

T 1 2 3 4 5 6 7 8 9 10
f(t) 2.5 3.5 4 3.5 6 10 9 6 6 2.4

a) 9.32
b) 8.32
c) 9.42
d) 9.52

Correct Answer: c.

Detailed Solution:

Using cubic B-spline function

 t3
 ; 0 ≤ t <1
 6
 3 2
 − 3t + 12t − 12t + 4 ; 1 ≤ t < 2
 6
B0, 4 (t ) =  3 2
3t − 24t + 60t − 44
 ; 2≤t <3
 6
  4 − t  3
  
; 3≤t < 4
 6
0; otherwise

Since we have to interpolate f(6.3), we can shift this function by 4.3 so that peak of cubic
function lie on 6.3. By doing this we get the contribution of known pixels for location t =
6.3.
The modified function can be written as,
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

 (t − 4.3)3
 ; 4.3 ≤ t < 5.3
 6
 3 2
 − 3(t − 4.3) + 12(t − 4.3) − 12(t − 4.3) + 4 ; 5.3 ≤ t < 6.3
 6
B4.3, 4 (t ) =  3(t − 4.3)3 − 24(t − 4.3)2 + 60(t − 4.3) − 44
 ; 6.3 ≤ t < 7.3
 6
  4 − (t − 4.3) 3
  
; 7.3 ≤ t < 8.3
 6
 0; otherwise
Substituting the values at t=5,6,7 and 8 (values are known at these locations) in proper
intervals, we get the coefficient (contribution) that has to be multiplied with intensity
values at location,
t= 5,6,7,8.
 (5 − 4.3)3
 ; 4.3 ≤ t < 5.3
 6
 3 2
 − 3(6 − 4.3) + 12(6 − 4.3) − 12(6 − 4.3) + 4 ; 5.3 ≤ t < 6.3
 6
B4.3, 4 (t ) =  (
3 7 − 4.3) − 24(7 − 4.3)2 + 60(7 − 4.3) − 44
3
 ; 6.3 ≤ t < 7.3
 6
  4 − (8 − 4.3)3
  
; 7.3 ≤ t < 8.3
 6
 0; otherwise
On solving this, we get
 0.0572; 4.3 ≤ t < 5.3

 0.592 ; 5.3 ≤ t < 6.3

B4.3, 4 (t ) =  0.3482; 6.3 ≤ t < 7.3

 0.0045 ; 7.3 ≤ t < 8.3
 0; otherwise
Now f (6.3) = ∑ pi Bi −4.3, 4 (t ) = 0.05728 * 6 + 0.592 *10 + 0.3482 * 9 + 0.0045 * 6 = 9.42

______________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 3:
For an image, the type of Interpolation where the intensity of the four neighboring pixels is used
to obtain intensity a new location is called __________________.

a) Bilinear interpolation
b) Cubic interpolation
c) Bicubic interpolation
d) Nearest neighbor interpolation

Correct Answer: a.

Detailed Solution:

Region of support for linear b-spline is two. But when we go for two dimensional interpolation
as in case of images, it will consider two pixels in horizontal direction and two pixels in vertical
direction.

______________________________________________________________________________

QUESTION 4:
Which of the following B-spline is not symmetric and hardly used?
a) Constant
b) Cubic
c) Linear
d) Quadratic

Correct Answer: d.

Detailed Solution:

Region of support for quadratic b-spline is 3 and it is asymmetric. Therefore, quadratic b-spline
is hardly used.
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 5:
A one-dimensional function f(x) is given below. What will be the value at location x = 5.5 using
modified constant interpolation.

X 1 2 3 4 5 6 7 8 9 10
f(x) 1.5 2.5 3 2.5 3 2.4 3 2.5 1 2.4

a) 2.43
b) 2.40
c) 2.34
d) 3.00

Correct Answer: b
Detailed Solution:

For constant interpolation, we substitute the unknown pixel value by nearest known pixel value.
For x=5.5, we consider B-Spline at x=6 since it includes 5.5 whereas B-Spline at x=5 excludes
x=5.5. Therefore f(5.5) = 2.40.
______________________________________________________________________________

QUESTION 6:

The type of Interpolation in which the intensity for an unknown location is found by assigning
the intensity of the nearest pixel is __________________.

a) Bicubic interpolation
b) Constant interpolation
c) Cubic interpolation
d) Bilinear interpolation

Correct Answer: b.

Detailed Solution:

Region of support for constant interpolation is one. So because of this it will look for a single
location which is nearest to it.
______________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 7:
What are the desirable properties of an interpolation function?
I. Finite Region of Support.
II. Smooth Interpolation.
III. Shift Invariant

a) Only I and II.

b) Only II and III.
c) None of I, II and III.
d) All I, II and III.

Correct Answer: d.
Detailed Solution:

Interpolation function must have finite region of support. Interpolation function should smooth
profile. Interpolation function must be shift invariant.
______________________________________________________________________________

QUESTION 8:
Zooming of an image can be viewed as __________________.
a) Over sampling
b) Under sampling
c) Critical sampling
d) Nyquist sampling
Correct Answer: a.

Detailed Solution:

Zooming of an image is basically estimating the sub-pixels and therefore can be viewed as
oversampling.
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 9:
What kind of sampling is required for scaling down (Shrinking) of an image?
a) Under sampling
b) Critical sampling
c) Over sampling
d) Nyquist sampling
Correct Answer: a
Detailed Solution:

Scaling down of an image is basically sampling less pixels, for example sampling every other
pixel will scale down image by 2. Therefore shrinking can be viewed as under sampling.

______________________________________________________________________________

QUESTION 10:
Which of the following B-spline function has maximum region of support?
a) Quadratic
b) Cubic
c) Linear
d) Constant

Correct Answer: b
Detailed Solution:

See the Class notes.

______________________________________________________________________________

************END*******
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Digital Image Processing

Assignment- Week 5
TYPE OF QUESTION: MCQ/MSQ
Number of questions: 10 Total mark: 10 X 1 = 10
______________________________________________________________________________

QUESTION 1:

Given two signals 𝑥𝑥[𝑛𝑛], 𝑦𝑦[𝑛𝑛] 𝑤𝑤ℎ𝑒𝑒𝑒𝑒𝑒𝑒 𝑛𝑛 ∈ {0,1,2, . . . 𝑁𝑁 − 1} and their DFT coefficients 𝑋𝑋[𝑘𝑘], 𝑌𝑌[𝑘𝑘]
𝑤𝑤ℎ𝑒𝑒𝑒𝑒𝑒𝑒 𝑘𝑘 ∈ {0,1,2, . . . 𝑁𝑁 − 1}. Taking inverse DFT of 𝑍𝑍[𝑘𝑘] = 𝑋𝑋[𝑘𝑘] × 𝑌𝑌[𝑘𝑘] is equivalent to

a) Linear convolution of 𝑥𝑥[𝑛𝑛]𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎[𝑛𝑛]

b) Circular/Cyclic convolution of 𝑥𝑥[𝑛𝑛] 𝑎𝑎𝑎𝑎𝑎𝑎 𝑦𝑦[𝑛𝑛]
c) Correlation of 𝑥𝑥[𝑛𝑛] 𝑎𝑎𝑎𝑎𝑎𝑎 𝑦𝑦[𝑛𝑛]
d) None of these
Correct Answer: b

Detailed Solution:

This is from basic DSP classes. See Class notes as well.

____________________________________________________________________________

QUESTION 2:
Determine the circular/cyclic convolution of two sequences x(n) = {1, 2, 3} and h(n) = {4, 5, 6}
a) y(n) = {9,12,10,4,1}
b) y(n) = {1,4,10,12,9}
c) y(n) = {31,31,28}
d) y(n) = {3,8,8}

Correct Answer: c

Detailed Solution:

Circular/cyclic convolution may be calculated as

1 3 2 4 31
Cicrular_Conv{x(n), h(n)} = �2 �
1 3 5� � = � 31�
3 2 1 6 28
______________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 3:
Which of the pair does form an Orthogonal Pair?
a. 𝑒𝑒 𝑗𝑗6𝜔𝜔𝜔𝜔 , 𝑒𝑒 𝑗𝑗2𝜔𝜔𝜔𝜔
b. 𝑠𝑠𝑠𝑠𝑠𝑠(2𝜋𝜋𝜋𝜋𝜋𝜋) , 𝑐𝑐os(4.5𝜋𝜋𝜋𝜋𝜋𝜋)
c. 𝑠𝑠𝑠𝑠𝑠𝑠(2.5𝜋𝜋𝜋𝜋𝜋𝜋) , 𝑐𝑐os(4𝜋𝜋𝜋𝜋𝜋𝜋)
d. All of the above
Correct Answer: a
Detailed Solution:

Only option a satisfy the orthogonality property of ∫𝑇𝑇 𝑓𝑓1 (𝑡𝑡) 𝑓𝑓2 (𝑡𝑡)𝑑𝑑𝑑𝑑 = 0 𝑎𝑎𝑎𝑎𝑎𝑎 ∫𝑇𝑇 𝑓𝑓1 (𝑡𝑡) 𝑓𝑓1 (𝑡𝑡)𝑑𝑑𝑑𝑑 =
𝑘𝑘 , ∫𝑇𝑇 𝑓𝑓2 (𝑡𝑡) 𝑓𝑓2 (𝑡𝑡)𝑑𝑑𝑑𝑑 = 𝑘𝑘

______________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 4:
For a given 4 × 4 image 𝑈𝑈 and 4 × 4 transformation matrix 𝐴𝐴, What will be the transformed
image 𝑉𝑉?

1 1 1 1 1 1 1 1
1 1 −1 1 −1 2 2 2 2
𝐴𝐴 = � � 𝑈𝑈 = � �
2 1 1 −1 −1 3 3 3 3
1 −1 −1 1 4 4 4 4

10 −2 −4 0
0 0 0 0
a) � �
0 0 0 0
0 0 0 0

10 0 0 0
−2 0 0 0
b) � �
−4 0 0 0
0 0 0 0

10 −2 −4 0
−2 0 0 0
c) � �
−4 0 0 0
0 0 0 0

d) None of the above

Correct Answer: b
Detailed Solution:

A two dimensional transformed is given as

𝑉𝑉 = 𝐴𝐴𝐴𝐴𝐴𝐴𝑇𝑇

1 1 1 1 1 1 1 1 1 1 1 1
1 1 −1 1 −1 2 2 2 2 1 −1 1 −1
= � �� �� �
4 1 1 −1 −1 3 3 3 3 1 1 −1 −1
1 −1 −1 1 4 4 4 4 1 −1 −1 1
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

10 0 0 0
−2 0 0 0
=� �
−4 0 0 0
0 0 0 0

______________________________________________________________________________

QUESTION 5:
How does image transformation represent a given image?
a. A series summation of a set of Symmetric Matrices.
b. A series summation of a set of Unitary Matrices
c. A series summation of a set of Skew-Symmetric Matrices.
d. None of the above.
Correct Answer: b
Detailed Solution:

Any image can be expressed as a scaled summation of basis images which are unitary matrices.
(Just like a 1-D signal can be expressed in terms of basis vectors).
______________________________________________________________________________

QUESTION 6:
Which of the following relation is true for one-dimension Fourier transformation?

∞
a. 𝐹𝐹(𝑢𝑢) = ∫−∞ 𝑓𝑓(𝑥𝑥)𝑒𝑒 𝑗𝑗2𝜋𝜋𝜋𝜋𝜋𝜋 𝑑𝑑𝑑𝑑
∞
b. 𝐹𝐹(𝑢𝑢) = ∫−∞ 𝑓𝑓(𝑥𝑥)𝑒𝑒 −𝑗𝑗2𝜋𝜋𝜋𝜋𝜋𝜋 𝑑𝑑𝑑𝑑
∞
c. 𝐹𝐹(𝑢𝑢) = ∫0 𝑓𝑓(𝑥𝑥)𝑒𝑒 −𝑗𝑗2𝜋𝜋𝜋𝜋𝜋𝜋 𝑑𝑑𝑑𝑑
d. All of the above

Correct Answer: b.

Detailed Solution:
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

From the definition of one-dimension Fourier transform,

∞
𝐹𝐹(𝑢𝑢) = � 𝑓𝑓(𝑥𝑥)𝑒𝑒 −𝑗𝑗2𝜋𝜋𝜋𝜋𝜋𝜋 𝑑𝑑𝑑𝑑
−∞

Hence option b is correct.

______________________________________________________________________________

QUESTION 7:
What will be the Kronecker product 𝐴𝐴 ⊗ 𝐵𝐵 , Given matrices A,B below?

1 2 2 1
𝐴𝐴 = � � , 𝐵𝐵 = � �
3 4 1 2

2 4 1 2
6 8 3 4
a) � �
1 2 2 4
3 4 6 8

2 1 4 2
1 2 2 4
b) � �
2 1 4 2
1 2 2 4

1 2 4 3
3 4 2 1
c) � �
2 1 3 4
4 3 1 2

d) None of the above.

Correct Answer: b
Detailed Solution:

Apply the Kronecker delta product formula where matrix A replaces individual elements of A
scaled by a value determined by matrix A.
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

______________________________________________________________________________

QUESTION 8:
What will be the basis image A11 for a given 4 × 4 transformation matrix 𝐴𝐴?

1 1 1 1
1 1 −1 1 −1
𝐴𝐴 = � �
2 1 1 −1 −1
1 −1 −1 1

1 1 1 1
1 1 1 1 1
a) � �
4 1 1 1 1
1 1 1 1

1 −1 1 −1
1 1 −1 1 −1
b) � �
4 1 −1 1 −1
1 −1 1 −1

−1 −1 −1 −1
1 1 1 1 1
c) � �
4 1 1 1 1
1 1 1 1

1 −1 1 −1
1 −1 1 −1 1
d) � �
4 1 −1 1 −1
−1 1 −1 1

Correct Answer: d
Detailed Solution:

1 1 1 1
1 1 −1 1 −1
Basis images are calculated using the columns of𝐴𝐴∗𝑇𝑇 = � �,
2 1 1 −1 −1
1 −1 −1 1
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Where 𝐴𝐴 is the transformation matrix.

1 1 −1 1 −1
1 −1 1−1 1 −1 1
𝐴𝐴11 = 𝑎𝑎1 𝑎𝑎1𝑇𝑇 = � � [1 −1 1 −1] = 4 � �
4 1 1 −1 1 −1
−1 −1 1 −1 1
______________________________________________________________________________

QUESTION 9:
This question has Statement 1 and Statement 2. Of the four choices given after the statements,
choose the one that best describes the two statements.
Statement1:The separability of orthogonal transformation matrix reduces the
computational complexity from 𝑂𝑂(𝑛𝑛4 ) 𝑡𝑡𝑡𝑡 𝑂𝑂(2𝑛𝑛3 )
Statement 2: Orthogonal transformation of Image can be thought of linear summation of
set of orthogonal basis images where coefficients of linear summation are estimated by
projecting image onto the orthogonal basis images.

a) Statement 1 is True and Statement 2 is False

b) Statement 1 is False and Statement 2 is True
c) Statement 1 is True and Statement 2 is True
d) Statement 1 is False and Statement 2 is False

Correct Answer: C
Detailed Solution:

Both the statements follow directly from lectures.

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 10:
Which of the following can be used as a transformation matrix?
1 1 1 1
1 1 𝜔𝜔1 𝜔𝜔2 𝜔𝜔3
a) 𝐴𝐴 = � � ; 𝜔𝜔1 = 𝑒𝑒 −𝑗𝑗2𝜋𝜋/4
2 1 𝜔𝜔2 𝜔𝜔4 𝜔𝜔6
1 𝜔𝜔3 𝜔𝜔6 𝜔𝜔9

 1 1 1 1 
 1 1 −1 −1 
1
b) A = 
2 2 − 2 0 0 
 
 0 0 − 2 2 

1 1 1 1
1 1 −1 1 −1
c) 𝐴𝐴 = � �
2 1 1 −1 −1
1 −1 −1 1

d) All of above

Correct Answer: d
Detailed Solution:

For unitary matrix, all the columns must be orthonormal or orthogonal. Matrix in option a is DFT
basis, c is Hadamard basis, All a, b, c satisfy orthonormality condition
______________________________________________________________________________

************END*******
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Digital Image Processing

Assignment- Week 6
TYPE OF QUESTION: MCQ/MSQ
Number of questions: 10 Total mark: 10 X 1 = 10
______________________________________________________________________________

QUESTION 1:

In the following ℱ{𝑓𝑓(𝑥𝑥, 𝑦𝑦)} means DFT of 𝑓𝑓(𝑥𝑥, 𝑦𝑦)

Which of the following the property is False for DFT?
a) ℱ{𝑎𝑎𝑎𝑎(𝑥𝑥, 𝑦𝑦)} = 𝑎𝑎ℱ{𝑓𝑓(𝑥𝑥, 𝑦𝑦)}
b) ℱ{𝑓𝑓(𝑥𝑥, 𝑦𝑦) + 𝑔𝑔(𝑥𝑥, 𝑦𝑦)} = ℱ{𝑓𝑓(𝑥𝑥, 𝑦𝑦)} + ℱ{𝑔𝑔(𝑥𝑥, 𝑦𝑦)}
c) ℱ{𝑓𝑓(𝑥𝑥, 𝑦𝑦). 𝑔𝑔(𝑥𝑥, 𝑦𝑦)} = ℱ{𝑓𝑓(𝑥𝑥, 𝑦𝑦)}. ℱ{𝑔𝑔(𝑥𝑥, 𝑦𝑦)}
2𝜋𝜋
d) ℱ{𝑓𝑓(𝑥𝑥 − 𝑥𝑥0 , 𝑦𝑦 − 𝑦𝑦0 )} = ℱ{𝑓𝑓(𝑥𝑥, 𝑦𝑦)}𝑒𝑒 − 𝑁𝑁 (𝑢𝑢𝑢𝑢0 +𝑣𝑣𝑣𝑣0)
Correct Answer: c
Detailed Solution:

Fourier transform is not distributive over multiplication. Please see class notes
____________________________________________________________________________

QUESTION 2:
Given that 𝑓𝑓(𝑥𝑥, 𝑦𝑦) 𝑎𝑎𝑎𝑎𝑎𝑎 𝑖𝑖𝑖𝑖𝑖𝑖 𝐷𝐷𝐷𝐷𝐷𝐷, 𝐹𝐹(𝑢𝑢, 𝑣𝑣)
Which of the following relation is used to display DFT coefficients properly.

2𝜋𝜋
a) 𝐹𝐹(𝑢𝑢 − 𝑢𝑢0 , 𝑣𝑣 − 𝑣𝑣0 ) ↔ 𝑓𝑓(𝑥𝑥, 𝑦𝑦)𝑒𝑒 − 𝑁𝑁 (𝑢𝑢0 𝑥𝑥+𝑣𝑣0 𝑦𝑦)
2𝜋𝜋
b) 𝑓𝑓(𝑥𝑥 − 𝑥𝑥0 , 𝑦𝑦 − 𝑦𝑦0 ) ↔ 𝐹𝐹(𝑢𝑢, 𝑣𝑣)𝑒𝑒 − 𝑁𝑁 (𝑢𝑢𝑢𝑢0 +𝑣𝑣𝑣𝑣0 )
c) 𝑎𝑎𝑎𝑎(𝑢𝑢, 𝑣𝑣) ↔ 𝑎𝑎𝑎𝑎(𝑥𝑥, 𝑦𝑦)
1 𝑢𝑢 𝑣𝑣
d)
|𝑎𝑎𝑎𝑎|
𝐹𝐹( , ) ↔ 𝑓𝑓(𝑎𝑎𝑎𝑎, 𝑏𝑏𝑏𝑏)
𝑎𝑎 𝑏𝑏
Correct Answer: a

Detailed Solution:
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Displaying DFT coefficients requires shiting in frequency domain, Hence option a .

______________________________________________________________________________

QUESTION 3:
Which of the following transformations preserves the signal strength (magnitude) of the data
a) Discrete Walsh Transform (DWT)
b) Discrete Fourier Transform (DFT)
c) KL Transform
d) All of the above
Correct Answer: d
Detailed Solution:

All unitary transforms preserve the signal strength (magnitude) of the data.
𝑦𝑦 = 𝐴𝐴𝐴𝐴; 𝐴𝐴 is transformation unitary matrix, x is data sample
Now Magnitude squared
‖𝑦𝑦‖22 = 𝑦𝑦 𝑇𝑇 𝑦𝑦 = (𝐴𝐴𝐴𝐴)𝑇𝑇 𝐴𝐴𝐴𝐴 = 𝑥𝑥 𝑇𝑇 𝐴𝐴𝑇𝑇 𝐴𝐴𝐴𝐴 = 𝑥𝑥 𝑇𝑇 𝐼𝐼𝐼𝐼 = 𝑥𝑥 𝑇𝑇 𝑥𝑥 = ‖𝑥𝑥‖22

For DWT and DFT consult lectures, KL transform specifies

𝑦𝑦 = 𝐴𝐴(𝑥𝑥 − 𝜇𝜇𝑥𝑥 ) ; Where matrix 𝐴𝐴 is formed by eigenvectors, such that first row of 𝐴𝐴 is
eigenvector corresponding to largest eigenvalue of 𝐶𝐶𝑥𝑥 = 𝐸𝐸[(𝑥𝑥 − 𝜇𝜇𝑥𝑥 )(𝑥𝑥 − 𝜇𝜇𝑥𝑥 )𝑇𝑇 ]; (𝐸𝐸[. ] Denotes
expectation), Second row of 𝐴𝐴 is eigenvector corresponding to second largest eigenvalue 𝐶𝐶𝑥𝑥
and so on.
To prove 𝐴𝐴 is a unitary matrix:
Notice that 𝐶𝐶𝑥𝑥 is a Symmetric Positive Semidefinite (PSD) matrix,
𝐶𝐶𝑥𝑥 = 𝐶𝐶𝑥𝑥𝑇𝑇 and eigenvalue of 𝐶𝐶𝑥𝑥 , 𝜆𝜆𝑖𝑖 ≥ 0
Proof of Symmetry:
𝐶𝐶𝑥𝑥𝑇𝑇 = (𝐸𝐸[(𝑥𝑥 − 𝜇𝜇𝑥𝑥 )(𝑥𝑥 − 𝜇𝜇𝑥𝑥 )𝑇𝑇 ])𝑇𝑇
𝑇𝑇
= 𝐸𝐸[�(𝑥𝑥 − 𝜇𝜇𝑥𝑥 )(𝑥𝑥 − 𝜇𝜇𝑥𝑥 )𝑇𝑇 � ]
= 𝐸𝐸[(𝑥𝑥 − 𝜇𝜇𝑥𝑥 )(𝑥𝑥 − 𝜇𝜇𝑥𝑥 )𝑇𝑇 ]
= 𝐶𝐶𝑥𝑥
As eigenvectors of real symmetric matrix are orthogonal,
Therefore 𝐴𝐴𝐴𝐴𝑇𝑇 = 𝐴𝐴𝑇𝑇 𝐴𝐴 = 𝐼𝐼
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 4:
What is the time complexity of Fast Fourier Transform for an N -point Discrete Fourier Transform?
a) 𝑂𝑂(𝑁𝑁 2 )
b) 𝑂𝑂(𝑁𝑁 log10 𝑁𝑁)
c) 𝑂𝑂(𝑁𝑁 log 2 𝑁𝑁)
d) None of the above
Correct Answer: c

Detailed Solution:

There is a total of 𝑀𝑀 = log 2 𝑁𝑁 stages of computation, each requiring 3/2𝑁𝑁 complex operations.
Hence, the total computational complexity is 𝑂𝑂(𝑁𝑁 log 2 𝑁𝑁).
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 5:
Which of the following corresponds to 8x8 Hadamard matrix?

1 1 1 1 1 1 1 1
⎡1 −1 1 −1 1 −1 1 −1⎤
⎢1 1 −1 −1 1 1 −1 −1⎥
⎢ ⎥
1 −1 −1 1 1 −1 −1 1⎥
a. ⎢
⎢1 1 1 1 −1 −1 −1 −1⎥
⎢1 −1 1 −1 −1 1 −1 1⎥
⎢1 1 −1 −1 −1 −1 1 1⎥
⎣1 −1 −1 1 −1 1 1 −1⎦

1 1 1 1 1 1 1 1
⎡1 −1 1 −1 1 −1 1 −1⎤
⎢1 1 −1 −1 1 1 −1 −1⎥
⎢ ⎥
1 −1 −1 −1 1 −1 −1 −1⎥
b. ⎢
⎢1 1 1 1 −1 −1 −1 −1⎥
⎢1 −1 1 −1 −1 1 −1 1⎥
⎢1 1 −1 −1 −1 −1 1 1⎥
⎣1 −1 −1 −1 −1 1 1 1⎦

1 1 1 1 1 1 1 1
⎡1 −1 1 −1 1 −1 1 −1⎤
⎢1 1 1 1 1 1 1 1⎥
⎢ ⎥
1 −1 1 −1 1 −1 1 −1⎥
c. ⎢
⎢1 1 1 1 1 1 1 1⎥
⎢1 −1 1 −1 1 −1 1 −1⎥
⎢1 1 1 1 1 1 1 1⎥
⎣1 −1 1 −1 1 −1 1 −1⎦

d. All of the above.

Correct Answer: a
Detailed Solution:

1 1 𝐻𝐻 𝐻𝐻𝑁𝑁
𝐻𝐻2 = � �and 𝐻𝐻2𝑁𝑁 = � 𝑁𝑁 �.
1 −1 𝐻𝐻𝑁𝑁 −𝐻𝐻𝑁𝑁
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

1 1 1 1 1 1 1 1
⎡1 −1 1 −1 1 −1 1 −1⎤
⎢
⎢1 1 −1 −1 1 1 −1 −1⎥⎥
Therefore, 𝐻𝐻8 = ⎢1 −1 −1 1 1 −1 −1 1 ⎥.
⎢1 1 1 1 −1 −1 −1 −1⎥
⎢1 −1 1 −1 −1 1 −1 1⎥
⎢1 1 −1 −1 −1 −1 1 1⎥
⎣1 −1 −1 1 −1 1 1 −1⎦

______________________________________________________________________________

QUESTION 6:

Which of the following recursive relation is used to obtain higher order Hadamared transform
kernels?

H N HN 
a. H 2 N = 
H N H N 
H N H N 
b. H 2 N =  
H N − H N 
 H − HN 
c. H 2 N =  N
− H N H N 
d. None of the above mentioned.
Correct Answer: b
Detailed Solution:

H HN 
H2N =  N is used to obtain higher order Hadamard transform kernels.
H N − H N 
______________________________________________________________________________

QUESTION 7:
Considering the cost and complexity of hardware required, which of the following transforms is
generally preferred for real-time applications?

a) Fast Walsh/Hadamard transform

b) DCT
c) FFT
d) KL Transform
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Correct Answer: a
Detailed Solution:

Hadamard/Walsh transform only need addition and subtraction operation on samples of

underlying signal, This is why they are preferred over FFT and DCT for low-cost, real-time
hardware applications.
______________________________________________________________________________

QUESTION 8:
Which of the following transforms have exactly same basis images as that of walsh transform
(ignore the order of basis images)?
a. Hadamard Transform
b. KL transform
c. DCT
d. Ordered Hadamard Transform
Correct Answer: d
Detailed Solution:

Ordered Hadamard Transform and walsh transform have same basis images, just in different
order

______________________________________________________________________________

QUESTION 9:
Imagine you have powerful servers with significant processing capabilities. You're looking for a
transform that can squeeze the most information out of data while efficiently representing it
with minimal loss. Which of the following transforms would be the best choice?

a) Fast Walsh/Hadamard transform

b) DCT
c) FFT
d) KL Transform
Correct Answer: d
Detailed Solution:
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

KL transform has highest energy compaction and representative power, thus when
computation is not an issue, KL transform is preferred over DCT and FFT for retrieval and
identification tasks

______________________________________________________________________________

QUESTION 10:
An application for compressing storage space of images uses CPU that has built it FPU - Floating
Point Unit (Hardware accelerated Sine, Cosine and exponentiation, floating numbers arithmetic,
etc.) which of the following transform is better suited for this application?

a) Fast Walsh/Hadamard transform

b) DCT
c) FFT
d) KL Transform
Correct Answer: b
Detailed Solution:

DCT is preferred for storage of images since it has high energy compaction while needing
moderate compute resources. Since the CPU has hardware accelerated Cosine function via.
FPU, DCT is most appropriate transform for this application.

KL transform has highest energy compaction, but it needs prohibitively high compute
requirements due to computation of eigenvectors ~𝑂𝑂(𝑛𝑛3 )

______________________________________________________________________________

************END*******
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Digital Image Processing

Assignment- Week 7
TYPE OF QUESTION: MCQ/MSQ
Number of questions: 10 Total mark: 10 X 1 = 10
______________________________________________________________________________

QUESTION 1:
Which of the following function cannot be a histogram equalization function.

a) f(x) = 3𝑥
b) f(x) = 5−𝑥
c) f(x) = 2x + 7
d) 𝑓(𝑥) = exp⁡(𝑥)

Correct Answer: b

Detailed Solution:

The function f () should be a monotonic increasing function.

________________________________________________________________

QUESTION 2:
Which process is used to correct the power-law response ?
a) Alpha Correction
b) Gamma correction
c) Luminance Correction
d) None of the above
Correct Answer: b

Detailed Explanation: Gamma correction is used correct the power-law response.

______________________________________________________________________________

QUESTION 3:
Log transformation is given by which formula

a) 𝑠 = 𝑐⁡𝑙𝑜𝑔(|𝑟| − 1), where c is the scaling constant and r is the input intensity.
b) 𝑠 = ⁡𝑙𝑜𝑔(1 − |𝑟|), where r is the input intensity.
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

c) 𝑠 = 𝑐⁡𝑙𝑜𝑔(|𝑟|), where c is the scaling constant and r is the input intensity.

d) 𝑠 = 𝑐⁡𝑙𝑜𝑔(1 + |𝑟|), where c is the scaling constant and r is the input intensity.

Correct Answer: d.

Detailed Explanation:⁡𝒔 = 𝒄⁡𝒍𝒐𝒈(𝟏 + |𝒓|) represents log transformation. Where c is the

scaling constant and r is the input intensity.

________________________________________________________________

QUESTION 4:
Which of the following is TRUE, for power law transformation?
𝑇(𝑟) = 𝑐𝑟 𝛾
Where, 𝑐 and 𝛾 are positive constants and r is the input intensity
a) 𝛾 < 1 =>Narrow range of dark input values is mapped into wider range of values
and wide range of high intensity values is mapped into narrow range of values

b) 𝛾 > 1 =>Narrow range of dark input values is mapped into wider range of values
and wide range of high intensity values is mapped into narrow range of values

c) 𝛾 < 1 =>Narrow range of dark input values is mapped into narrow range of values
and wide range of high intensity values is mapped into wider range of values

d) None of the above.

Correct Answer: a.

Detailed Solution:

Using the gamma curve, we can say that 𝛾 < 1 =>Narrow range of dark input values is mapped
into wider range of values and wide range of high intensity values is mapped into narrow range
of values

______________________________________________________________________________

QUESTION 5:
For decoding a K L Transformed image at the receiver side, we require
a) Transformed image
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

b) Transformation matrix
c) Original Image
d) Both Transformed image and transformation matrix.
Correct Answer: d
Detailed Explanation: For decoding KL transformed image, we need both the transformed
image and transformation matrix at the decoder side.
______________________________________________________________________________

QUESTION 6:
Which of the following statement is (are) TRUE for KL Transform?

I. No fast computational algorithms are available for its implementation.

II. Transformation matrix is image independent.
III. KL Transform is optimum in the sense that it minimizes mean square error
between 𝒙 and 𝒙̂, where 𝒙 is the original vector and 𝒙 ̂ is the reconstructed
vector.
IV. Only transformed matrix is sufficient for inverse K-L transform at receiver.

a) Only I
b) Only I, III
c) Only I, II, IV
d) Only I, II and III

Correct Answer: b.
Detailed Solution:

KL Transform kernel minimizes mean square error between 𝒙 and 𝒙 ̂, where 𝒙 is the original
̂ is the reconstructed vector. Also the transformation matrix uses statistical properties
vector and 𝒙
of data and therefore it is not fixed and requires to be transmitted to the receiver for inverse
transformation. No fast computational algorithms are available for its implementation.

Hence option b is correct.

_____________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 7:

For the following vectors, find the transformation matrix A, for K-L transform
 1  3   3   5
 ,  ,   and  ,
 1  0   2  1

1 1 1 
a) A   
2 1  1
11  1
b) A  1 1 

2 
1 0
c) A   
0 1 
1 3 1 
d) A   
10 1  1

Correct Answer: c.

Detailed Solution:

1 12  3

4  4  1
Mean vector of the data =

  2  4 0
M 1    2 0   
0 0 0

0 0 0 
M 2   0  1   
 1 0 1 

0  0 0 
M 3   0 1   
1 0 1 

 2  4 0
M 4   2 0   
0  0 0
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

M1  M 2  M 3  M 4 2 0 
C  
4 0 0.5
Using properties of eigen values,

1  2  2.5 And 1 * 2  1 , on solving this we get,

1  2 and 2  0.5
Now for Eigen vectors CX   X

2 0   x1  x 
For 1  2 ,      2 1  , on solving for x1 and x2 , we get x2 = 0. So eigen
0 0.5  x2   x2 
vector can be
 x1  1
 x   0
 2  
Similarly for 2  0.5 , eigen vector can be

 x1  0
 x   1
 2  

Now we arrange these vectors as rows in decreasing values of corresponding eigen values to get
transformation kernel.

1 0
A 
0 1 

Hence option c is correct.

______________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 8:

Which of the following can result in forming low contrast image?

a) Poor illumination of the scene to be photographed.
b) Lack of dynamic range of imaging sensor used for image aquisition.
c) Wrong setting of lens aperture during image acquisition.
d) All of the above

Correct Answer: d.

Detailed Solution:

All the given condition can result in low contrast image.

Hence option d is correct.

_______________________________________________________________________

QUESTION 9:

For the following vectors, find the transformation matrix A, for K-L transform.
 1   2   3  4  5
  ,   ,   ,   and   .
 6  7  8  9  10 

a) 𝐴 = ⁡ (1 1
)
1 −1

b) 𝐴 = ⁡ (1 −1)
1 −1

c) 𝐴 = ⁡ ( 1 1
)
−1 −1

d) All of the above

Correct Answer: a.
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Detailed Solution:

1 15  3

5  40  8
Mean vector of data =

 2   4 4
M 1     2 2   
 2   4 4

 1 1 1
M 2     1 1   
 1 1 1

0  0 0 
M 3     0 0   
0  0 0 

1 1 1
M 4    1 1   
1 1 1

 2  4 4
M 5     2 2   
 2  4 4

M1  M 2  M 3  M 4  M 5 2 2
C  
5 2 2

Using properties of eigen values,

1  2  4 And 1.2  0 , on solving this we get,

1  4 and 2  0

Now for Eigen vectors CX   X

 2 2  x1  x 
For   4 ,      4  1  , on solving for x1 and x2 , we get
 2 2  x2   x2 
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

 x1  1
 x   1
 2  

 2 2  x1  x 
For   0 ,      0  1  , on solving for x1 and x2 , we get
 2 2  x2   x2 

 x1   1 
 x    1
 2  

Now we arrange these vectors as rows in decreasing values of corresponding eigen values to get
transformation kernel.
1 1
𝐴 =⁡( )
1 −1

Hence option a is correct.

______________________________________________________________________________

QUESTION 10:

Which of the following are true about Image Negative in image enhancement operation?

i. Image Negative operation produces image equivalent to a photographic

negative.
ii. It is used to increase the dynamic range of gray levels images.
iii. S = T(r) = L-r -1.

a) Only I
b) Only II
c) Only I and III
d) I, II and III

Correct Answer: c.

Detailed Solution:
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Both I and II are true regarding image negative.

Hence option c is correct.

________________________________________________________________

************END*******
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Digital Image Processing

Assignment- Week 8
TYPE OF QUESTION: MCQ/MSQ
Number of questions: 10 Total mark: 10 X 1 = 10
______________________________________________________________________________

QUESTION 1:

Laplacian operator can be used for which of the following operations?

a) Contrast Adjustment

b) Edge detection

c) Image Averaging

d) All of the above

Correct Answer: b.

Detailed Solution:

Laplacian operator is a second order difference operator and therefore can be used in edge
detection.

Hence option b is correct.

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 2:

Which of the following is a first-order derivative operator?

a) Laplacian of Gaussian

b) Laplacian

c) Sobel

d) All of the above

Correct Answer: c.

Detailed Solution:

  − 1 0 1 
 
Sobel operator  − 2 0 2  is first order difference.
 
  − 1 0 1 
 

Hence option c is correct.

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 3:

The images shown below (a & b) are quite different, but their histograms are the same. Suppose
that each image is blurred with a 3x3 averaging mask. Which of the following is (are) true?

a) Histogram of the blurred images will remain same.

b) Histograms of the blurred images will be different from each other.

c) Histograms of the blurred images will change but they will be identical.

d) None of the above.

Correct Answer: b.
Detailed Solution: Figure (b) has more edge pixels than figure (a). So after blurring, there
are more pixels in figure (b) whose intensity values will change in comparison to that of
figure (a). Therefore, histograms of the blurred images will be different from each other.
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 4:

Which of the following can be a used as a smoothing filter?

1 1 1
a) 1 − 8 1
1 1 1
1 2 1 
b) 2 4 2
1 2 1 
0 1 0 
1
c) 1 2 1
6
0 1 0
d) None of the above.

Correct Answer: c.

Detailed Solution:

1 1 1
1 − 8 1 is suitable for edge detection. Difference operator is present.
 
1 1 1

1 2 1 
2 4 2 is not normalized properly.
 
1 2 1 
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

0 1 0 
1
 1 2 1 is suitable for smoothing. Hence option c is correct.
6
0 1 0

______________________________________________________________________________

QUESTION 5:

Which of the following is true for Laplacian filter?

a) Zero along ramp of constant slope.

b) Robust to noise.

c) Helps in finding appropriate location of edge.

d) Both a and c.

Correct Answer: d.

Detailed Solution:

Laplacian is a second order derivative filter.

Second order derivative will have zero response along ramp of constant slope.

Because of second order derivative, noise is also amplified.

Zero crossing of double derivative provides most appropriate location of an edge.

Hence option d is correct.

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 6:
An image has the gray level Probability Distribution Function (PDF - or gray level histogram
normalised by number of pixels) of Pr(r) shown below.

Find the pixel transformation y = g(r) such that after transformation the image has a flat PDF, i.e.
which accomplishes histogram equalisation. Assume continuous variables r,y.
a. 2 − 𝑟𝑟
b. 2 − 2𝑟𝑟
c. 2𝑟𝑟 − 𝑟𝑟 2
d. −𝑟𝑟 2
Correct Answer: c
Detailed Solution:
We know the required transfer function is the cumulative probability distribution of Pr(r).
Here, Pr(𝑟𝑟) = 2 − 2𝑟𝑟
𝑟𝑟 1
Hence, 𝑦𝑦 = 𝑔𝑔(𝑟𝑟) = ∫0 𝑝𝑝𝜏𝜏 (𝜏𝜏). 𝑑𝑑𝑑𝑑/ ∫0 𝑝𝑝𝜏𝜏 (𝜏𝜏). 𝑑𝑑𝑑𝑑
So, y(𝑟𝑟) = 2𝑟𝑟 − 𝑟𝑟 2
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 7:

Which of the filter can be used to obtain horizontal edges in an image?

1 2 1

a)  0 0 0 
− 1 − 2 − 1

1 0 − 1 
b) 2 0 − 2
1 0 − 1 

c) Both a and b

d) None of the above

Correct Answer: a.

Detailed Solution:

1 2 1
0
 0 0  is vertical Sobel operator which does a averaging in horizontal direction
− 1 − 2 − 1
and difference operator in vertical direction. Therefore it can be used to detect horizontal
edge in an image.

Hence option a is correct.

______________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 8:
Do histogram equalisation on the following image which has 6 discrete pixel levels (0 - 5),
transforming it into a histogram equalised image also with 6 discrete grey levels in the range (0-
5).

4 1 3 2
3 1 1 1
� �
0 1 5 2
1 1 2 2

a) 5 3 4 4
4 3 3 3
� �
1 3 5 4
3 3 4 4

b) 3 3 5 4
4 3 2 3
� �
3 3 2 4
3 1 1 1
c) 4 3 2 2
5 3 3 1
� �
1 3 5 4
3 3 5 5
d) None of the above

Correct Answer: b.

Detailed Solution:

Let r represents the intensity levels, pr (r ) represents the probability of occurrence of r th

r
intensity level. s k = T (r ) = ∑ p r (i ) represents the transformation function from r to s .
i =0
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

𝑠𝑠−𝑠𝑠
𝑠𝑠 ′ = 𝐼𝐼𝐼𝐼𝐼𝐼 ��𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑚𝑚𝑚𝑚𝑚𝑚 ()� ∗ 5 + 0.5�
𝑚𝑚𝑚𝑚𝑚𝑚

5 3 4 4
4 3 3 3
Using transformation from above table, we get the equalized image as � �
1 3 5 4
3 3 4 4
Hence option a is correct.
_____________________________________________________________________________

QUESTION 9:
Consider an 8-bit grayscale image with 256 possible gray levels (0 to 255). If a certain gray level
'i' in the image has a normalized CDF of 0.5, what is the corresponding equalized gray level?

a. 127
b. 128
c. 129
d. 130

Correct Answer: b
Detailed Solution:
normalized CDF of 0.5

New gray level = round(0.5×255) = round(127.5)=128

Answer: B
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 10:

Which of the following is/are true for histogram of an image?

I. It represents the distribution of intensity values in an image.
II. It represents global appearance of an image.
III. Two different images may have same histogram.

a) Only I
b) Only I and II
c) Only I and III
d) All I, II and III

Correct Answer: d.

Detailed Solution:

Histogram is basically the probability distribution of intensity values in an image. It does not
capture any spatial correlation of intensity values but it captures the global appearance of an
image. Also two images may have same histogram even if they look different i.e. having
different local appearance but because of same distribution of intensity values considering the
entire image.

Hence option d is correct.

____________________________________________________________________

************END*******
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Digital Image Processing

Assignment- Week 9
TYPE OF QUESTION: MCQ/MSQ
Number of questions: 10 Total mark: 10 X 1 = 10
______________________________________________________________________________

QUESTION 1:

In which of the following filter, the intensity at a particular point in the image, is a product of two
terms, one is the illumination term, other one is the reflectance term ?
a) Laplacian filter

b) High Boost filter

c) Gaussian filter

d) Homomorphic filter

Correct Answer: d.
Detailed Solution:

In homomorphic filter, the intensity at a particular point in the image, is a product of two terms,
one is the illumination term, another one is the reflectance term is assumed.

Hence option d is correct.

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 2:

Which of the following filter does not produce ringing effect?

I. Gaussian Low Pass Filter
II. Ideal low pass filter
III. Low Pass Butterworth Filter of higher order
IV. Low Pass Butterworth Filter of low order

a) I, II

b) I, III

c) I, IV

d) I, II, III

Correct Answer: c.

Detailed Solution:

Since there is no abrupt change in frequency response of Gaussian low pass filter and low order
Butterworth filter, hence no ringing effect.

Hence option c is correct.

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 3:

1
The following equation corresponds to a filter, H(u, v) = D 2n
1+ � o �
D(u,v)

Where, n is a positive integer, Do is cut-off frequency, which of the following is true?

a) There is a ringing effect for higher value of n

b) There is no ringing effect for higher value of n

c) Ringing effect in independent of n

d) None of these

Correct Answer: a.

Detailed Solution:

Given filter characteristics represents Butterworth low pass filter. Higher order Butterworth
characteristics approaches to ideal low pass filter and there is abrupt change in frequency response,
hence ringing effect is observed.

Hence option a is correct.

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 4:
In case of inverse filtering, if the cut off frequency is significantly high then which of the following
is true?
a) Restored image becomes blurry

b) Restored image close to original image

c) Noise in restored image dominates

d) None of these
Correct Answer: c.
Detailed Solution: The reconstructed image in inverse filtering is given as
N (u, v)
Fˆ (u , v) = F (u, v) + , where F (u , v) is original image, N (u , v) is noise and H (u , v) is the
H (u, v)
degradation function.
It is assumed the image is low frequency signal and noise is high frequency signal, now if cut off
N (u , v)
frequency is significantly high then becomes higher and Noise in restored image
H (u , v)
dominates

______________________________________________________________________________

QUESTION 5:
A function is linear position invariant if it satisfies _________
a) Superposition theorem
b) Position invariance
c) Both a and b
d) None of the above mentioned

Correct Answer: c.
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Detailed Solution:

A function is linear position invariant if it satisfies superposition principle and position invariance.

Hence option c is correct

______________________________________________________________________________

QUESTION 6:

Which of the following function satisfy linearity property?

a. 𝑥𝑥 = 𝑡𝑡2 + 5
b. 𝑥𝑥 = 2𝑡𝑡 3 + 3
𝑑𝑑𝑑𝑑(𝑧𝑧)
c. 𝑥𝑥(𝑧𝑧) =
𝑑𝑑𝑑𝑑
d. All of above

Correct Answer: c

Detailed Explanation: Differentiator function satisfy both additivity and homogeneity

property.

______________________________________________________________________________

QUESTION 7:

Which of the following function satisfy linear position invariance property?

a. 𝑦𝑦(𝑛𝑛) = 𝑛𝑛 𝑥𝑥(𝑛𝑛)
b. 𝑦𝑦(𝑛𝑛) = 𝑥𝑥 2 (𝑛𝑛)
c. 𝑦𝑦(𝑛𝑛) = 0.5(𝑥𝑥(𝑛𝑛) + 𝑥𝑥(𝑛𝑛 − 1))
d. All of above

Correct Answer: c

Detailed Explanation: Averaging function satisfy both additivity and homogeneity property.
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 8:

In which of the following method estimation of degradation function is required?

a) Image restoration

b) Image enhancement

c) Both a and b

d) None of the above

Correct Answer: a.

Detailed Solution:

In case of image enhancement or simple noise filtering, we do not make use of any of the
degradation model or we do not bother about what is the process which is degrading the image.
Whereas in case of image restoration, we talk about the degradation model, we try to estimate the
model that has degraded the image and using that model we apply the inverse process and try to
restore the image.

Hence option a is correct.

_____________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 9:

Which of the following corresponds to Gaussian low pass filter?

− D 2 ( u ,v ) / 2 Do2
a) H (u , v) = e

− D 2 ( u ,v ) / 2 Do2
b) H (u , v) = 1 − e

1 if D(u,v) ≤ D0
c) H (u , v) = 
0 if D(u,v) > D0

d) None of these

Correct Answer: a.

Detailed Solution:

− D 2 ( u ,v ) / 2 Do2
Gaussian low pass filter is given as H (u , v) = e .

Hence option a is correct.

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 10:

Which of the following statements is (are) true?

Statement 1: f(x, y) cannot be reconstructed by using f(x, y) = H −1 (g(x, y) − η(x, y))
in degradation model because it requires a solution to a large number of simultaneous
linear equations.

Statement 2: Actual degradation function is seldom known completely during image

restoration.
Statement 3: Mathematical modeling for degradation cannot take into account
environmental conditions that cause degradation.

a) Statement Statement 2: Statement 3:

1: True True True
b) Statement Statement 2: Statement 3:
1: True True False
c) Statement Statement 2: Statement 3:
1: False True False
d) Statement Statement 2: Statement 3:
1: False False True

Correct Answer: b.

Detailed Solution:
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

For solving exactly f(x, y) = H −1 (g(x, y) − η(x, y)) and the number of variables would very large
(of the size of the image) and difficult to reconstruct using this relation. Consequently, in image
restoration, we never know the exact degradation function.

Appropriate mathematical modeling for degradation can take into account environmental
conditions that cause degradation.

Hence option b is correct.

____________________________________________________________________

************END*******
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Digital Image Processing

Assignment- Week 10
TYPE OF QUESTION: MCQ/MSQ
Number of questions: 10 Total mark: 10 X 1 = 10
______________________________________________________________________________

QUESTION 1:

Which of the following statement(s) is (are) true?

Statement 1: The drawback of Wiener filter is that constant approximation of a ratio of power
spectra is not always suitable
Statement 2: In constraint least square filtering using Iterative Selection of Gamma requires
knowledge of mean and variance of noise

a) Statement 1 is True; Statement 2 is False

b) Statement 1 is True; Statement 2 is True

c) Statement 1 is False; Statement 2 is True

d) Statement 1 is False; Statement 2 is False

Correct Answer: b.

Detailed Solution:

In Wiener filter is that constant approximation of a ratio of power spectra is not always suitable.

Constraint least square filtering requires noise parameter estimation for iterative search of
Gamma.

Hence option b is correct.

_______________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 2:
Which of the following the most suitable metric for match or mismatch measure in template
matching?
a) Sum of squared difference.

b) Cross co-relation

c) Normalised cross-correlation

d) None of these

Correct Answer: c.

Detailed Solution:

Normalized cross correlation is the most suitable metric for match or mismatch measure in
template matching as it normalizes the correlation with the matching image window pixels.

Hence option c is correct.

______________________________________________________________________________

QUESTION 3:
In which of the following application(s), image registration is used.
a) Template Matching

b) Mosaicing

c) Image Fusion

d) All of these

Correct Answer: d.
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Detailed Solution:

Template matching, mosaicking, image fusion are all application of image registration as
these requires some sort of matching and alignment between images.

Hence option d is correct.

______________________________________________________________________________

QUESTION 4:

In the minimum error function given by the expression

 
 1 H (u , v
2 
Fˆ (u , v) =  G (u , v)
 H (u , v) H (u , v 2 + S n (u , v) 
 S f (u , v) 

Where, 𝑆𝑆𝑓𝑓 (𝑢𝑢, 𝑣𝑣) is Power Spectrum of Original Image, 𝑆𝑆𝑛𝑛 (𝑢𝑢, 𝑣𝑣) is Noise Power
Spectrum, 𝐻𝐻(𝑢𝑢, 𝑣𝑣) is degradation function. The Wiener Filter is equal to the Inverse filter when

a) 𝑆𝑆𝑛𝑛 (𝑢𝑢, 𝑣𝑣) = 1

b) 𝑆𝑆𝑛𝑛 (𝑢𝑢, 𝑣𝑣) = 0

c) 𝑆𝑆𝑛𝑛 (𝑢𝑢, 𝑣𝑣) = 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 (except 0, 1)

d) None of the above

Correct Answer: b.

Detailed Solution:

If we substitute 𝑆𝑆𝑛𝑛 (𝑢𝑢, 𝑣𝑣) = 0 in the following relation

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

 
2
∧  1 H ( u, v ) 
F ( u, v ) =  .  G ( u , v ) , then it becomes
 H ( u , v ) H ( u , v ) 2 + Sη (u , v) 

 Sf ( u , v ) 
∧  1 
F ( u, v ) =   G ( u , v ) , which is nothing but inverse filtering.
 H ( u, v ) 
Hence option b is correct.
______________________________________________________________________________

QUESTION 5:

In which of the following restoration techniques, noise parameters estimation is required?

a) Constrained least square filtering

b) Weiner filtering

c) Inverse filtering

d) All of these

Correct Answer: a.

Detailed Solution:

Noise parameter estimation is required for image restoration using constrained least square
filtering.

Hence option a is correct.

______________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 6:
In Minimum Mean Square Error Filtering, which of the following is/are true?

I. Both the degradation function and statistical characteristics of noise are considered for
restoration.

II. Both the image and noise are considered as random processes and objective is to find an
estimate.

a) Only I
b) Only II
c) Both I and II
d) None of the above mentioned.

Correct Answer: c.

Detailed Solution:

In Minimum Mean Square Error Filtering,

Both the degradation function and statistical characteristics of noise are considered for
restoration.

Both the image and noise are considered as random processes and objective is to find an
estimate.

Hence option c is correct.

_
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

_____________________________________________________________________________

QUESTION 7:
The advantage of the Wiener filter over Inverse Filter is due to which of the following options?

a) Transfer function of Wiener Filter is chosen to minimise the mean square error
using statistical information of only image fields.
b) Transfer function of Wiener Filter is chosen to minimise the mean square error
using statistical information of only noise fields.
c) Transfer function of Wiener Filter is chosen to minimise the mean square error
using statistical information of both image and noise fields.
d) None of the above

Correct Answer: c.

Detailed Solution:

Transfer function of Wiener Filter is chosen to minimise the mean square error using statistical
information of both image and noise fields.

Hence option c is correct.

______________________________________________________________________________

QUESTION 8:

For image registration of two images, minimum point correspondences are required to establish
pixel relations of an object?

a) 8

b) 2

c) 6

d) 4
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Correct Answer: d.

Detailed Solution:

Let (x', y') be pixel coordinates in one image and (x, y) pixel coordinates in other image. These are
related as x' = k1x+k2y+k3xy+k4 and y' = k4x+k6y+k7xy+k8.

We have 8 unknowns and each correspondence pair gives us two equations. So we require at least
4 point correspondence.

Hence option d is correct.

______________________________________________________________________________

QUESTION 9:

Which of the following filter can be used for remove noise from image given below? (Noisy
image given in the left side and original noise free image is given in the right side.)

a) Low pass filter

b) Band pass filter
c) High pass filter
d) Band reject filter

Correct Answer: d.
Detailed Solution:
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

The given image is corrupted via periodic noise. In case of periodic noise, we have
significantly higher response at frequencies equal to the frequency of the periodic noise and
these can be eliminated with the help of a band reject filter.

______________________________________________________________________________

QUESTION 10:
Assume the degradation model for atmospheric turbulence is defined in frequency domain as
2 2
H (u , v) = e − K (u +v )5 / 6 . If the value of K is decreased, then atmospheric turbulence __________.

a) Decreases

b) Remains constant

c) Increases

d) None of these

Correct Answer: a.

Detailed Solution:

If the value of K is decreased, then atmospheric turbulence also decreases.

If K= 0 H ( u,v ) = 1 , that means input and output is same and no degradation

Hence option a is correct.

______________________________________________________________________________

************END*******
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Digital Image Processing

Assignment- Week 11
TYPE OF QUESTION: MCQ/MSQ
Number of questions: 10 Total mark: 10 X 1 = 10
______________________________________________________________________________

QUESTION 1:

If Red, Green, and Blue have values 0, 150 and 150 respectively then its corresponding values in
HSI are

a. H = π , S = 0 and I = 100

b. H = 0 , S = 1 and I = 100

c. H = π , S = 1 and I = 100

d. H = π , S = 0.5 and I = 100

Correct Answer: c.

Detailed Solution:

Using the transformation formula of RGB to HSI model.

θ if B ≤ G  ( R − G ) + ( R − B) 
H = , where θ = cos −1  ,
 360 − θ if B > G  2 ( R − G ) 2 + ( R − B)(G − B) 
 

 (0 − 150) + (0 − 150) 
θ = cos −1   = π; H = π
 2 (0 − 150) 2 + (0 − 150)(150 − 150) 
 

3
S = 1− [min( R, G, B)] = 1 − 0 = 1
R+G+ B 300
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

R + G + B 0 + 150 + 150
I= = = 100
3 3
Hence option c is correct.
____________________________________________________________________________

QUESTION 2:

Where do we find the spectrum colors in chromaticity diagram?

a. Inside

b. Outside

c. On boundaries

d. All of the above

Correct Answer: c.

Detailed Solution:

Following chromaticity diagram, spectrum colors are present at boundaries because these are pure
colors.
Hence option c is correct.
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

______________________________________________________________________________

QUESTION 3:

Match the range of parameters of HSI model?

I. Hue (i) 0 - 2π
II. Saturation (ii) 0 – 255
III. Intensity (iii) 0 – 1

a) I.- (ii), II- (i), III.- (iii)

b) I.- (i), II- (iii), III.- (ii)

c) I.- (iii), II- (i), III.- (ii)

d) None of these

Correct Answer: b.

Detailed Solution:

Hue is angle measured from the red color axis in anticlockwise direction, in the color hexagon and
it lies in the range 0 to 2π. Intensity ranges from 0 to 255 and saturation can be 0 to 1 or 1 to 255
depending on the normalization.

Hence option b is correct.

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

______________________________________________________________________________

QUESTION 4:

Which of the following color models is suitable from perception point of view?

a) RGB

b) CMYK

c) HSI

d) All of the above mentioned.

Correct Answer: c.

Detailed Solution:

HSI is suitable color space from perception point of view.

Hence option c is correct.

______________________________________________________________________________

QUESTION 5:
Range of saturation in HSI model is _________________

a. 0 – 255
b. 0 – 2π
c. 0–π
d. 0-1

Correct Answer: d.

Detailed Solution:
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Using the RGB to HSI color conversion relation, for saturation we have

3
S = 1− [min( R, G, B)],
R+G+ B
It is maximum when at least one R, G and B is zero, then S=1.

It is minimum when R, G and B, all are equal to 255, then S=0.

Hence option d is correct.

______________________________________________________________________________

QUESTION 6:

If Red, Green, and Blue have values 100, 50 and 100 respectively then its corresponding
normalized values in CMY are

a. C = 0.4, M = 0.5 and Y = 0.5

b. C = 0.6, M = 0.8 and Y = 0.6

c. C = 0.5, M = 0.5 and Y = 0.4

d. C = 0.4, M = 0.5 and Y = 0.4.

Correct Answer: b

Detailed Solution:

Normalized R = 0.4, G = 0.2 and B = 0.4.

C  1  R 
 M=  1 − G 
     
 Y  1  B  𝐶𝐶 1 0.4
, �𝑀𝑀� = �1� − �0.2�,
𝑌𝑌 1 0.4
C = 0.6, M = 0.8 and Y= 0.6
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Hence option b is correct.

______________________________________________________________________________

QUESTION 7:

Which of the following expression is correct for conversion of RGB values to saturation of HSI
model?

3
a) S = 1 − [max(R, G, B)]
R+G+ B
3
b) S = 1 + [min( R, G, B)]
R+G+ B
3
c) S = 1 − [min( R, G, B)]
R+G + B
d) None of these

Correct Answer: c.

Detailed Solution:

Using the RGB to HSI color conversion relation, for saturation we have

3
S = 1− [min( R, G, B)]
R+G + B

Hence option c is correct.

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

______________________________________________________________________________

QUESTION 8:
In the RGB space given below, which of following point indicates white color?

a. A
b. B
c. C
d. D

Correct Answer: D

Detailed Explanation: White point is point D of the plane (1, 1, 1).

______________________________________________________________________________

QUESTION 9:

If Red, Green, and Blue have values 150, 100 and 50 respectively then its corresponding values
in HSI are

a. H = 30o , S = 0.5 and I = 100

b. H = 0 , S = 1 and I = 85

c. H = 60o , S = 0.5 and I = 150

d. H = 0 , S = 0.5 and I = 100

Correct Answer: a

Detailed Solution:
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Using the transformation formula of RGB to HSI model.

θ if B ≤ G  ( R − G ) + ( R − B) 
H = , where θ = cos −1  ,
 360 − θ if B > G  2 ( R − G ) 2 + ( R − B)(G − B) 
 
3
𝑆𝑆 = 1 − [𝑚𝑚𝑚𝑚𝑚𝑚( 𝑅𝑅, 𝐺𝐺, 𝐵𝐵)]
𝑅𝑅 + 𝐺𝐺 + 𝐵𝐵

𝑅𝑅 + 𝐺𝐺 + 𝐵𝐵
𝐼𝐼 =
3
Hence option a is correct.

______________________________________________________________________________

QUESTION 10:

What are pigment primary colours?

a. Red, Green and Blue

b. Magenta, Cyan and Yellow
c. None of the above
d. Both (a) and (b)

Correct Answer: b.

Detailed Solution:

Magenta, Cyan and Yellow are pigment primary colours.

Hence option b is correct.

______________________________________________________________________________

************END*******
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Digital Image Processing

Assignment- Week 12
TYPE OF QUESTION: MCQ/MSQ

Number of questions: 10 Total mark: 10 X 1 = 10

______________________________________________________________________________

QUESTION 1:

Which of the following second order operator is most robust to noise in edge filtering?

a. Sobel operator

b. Laplacian operator

c. Prewitt operator

d. Laplacian of Gaussian operator

Correct Answer: d.

Detailed Solution:

In case of Laplacian of Gaussian operator, it is basically similar to performing low pass Gaussian
and the applying Laplacian operator. So this low pass operation minimize the noise and makes
LOG robust to noise in edge filtering.
Hence option d is correct.
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

____________________________________________________________________________

QUESTION 2:

Which of the following is (are) Sobel operator?

A
−1 0 1
�−2 0 2�
−1 0 1
B
1 1 1
�0 0 0�
−1 −1 −1
C
−1 0 1
�−1 0 1�
−1 0 1
D
1 2 1
�0 0 0�
−1 2 −1

a. Only A

b. Only C

c. Both A and B

d. Both A and D

Correct Answer: d.

Detailed Solution:
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

−1 0 1
�−2 0 2� corresponds to horizontal sobel filter and
−1 0 1

1 2 1
�0 0 0 � corresponds to vertical sobel filter.
−1 −2 −1

Hence option d is correct.

______________________________________________________________________________

QUESTION 3:
In a given image strip (S), assuming necessary zero padding find second order derivative

S= 4 4 4 4 4 4 0 0 0 0

a) − 4 0 0 0 0 − 4 4 0 0 0

b) 4 0 0 0 0 4 − 4 0 0 0

c) − 4 0 0 0 0 4 4 0 0 0

d) 4 0 0 0 0 4 4 0 0 0

Correct Answer: a.

Detailed Solution:

Using the relation, f " ( x)= f ( x + 1) + f ( x − 1) − 2 f ( x) . For example for f ( 0 ) ,

=f ( 0 ) f (1) + f ( −1) − 2 f ( 0 ) , here f ( −1) = 0 , therefore f ( 0 ) =4 + 0 − 2* 4 =−4 .

Similarly, f (1) = f ( 2 ) + f ( 0 ) − 2 f (1) = 4 + 4 − 2* 4 = 0 .

In this way we can compute the second order derivative for this sequence.
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Hence option a is correct.

______________________________________________________________________________

QUESTION 4:

 − 1 0 1
Which of the following is (are) are true for horizontal Sobel operator, Gx = − 2 0 2 ?
 − 1 0 1 

a. Perform averaging in vertical direction and differentiation in horizontal direction

b. Perform averaging in horizontal direction and differentiation in vertical direction

c. It can be used for detection vertical edges

d. Both a and c

Correct Answer: a.

Detailed Solution:

1 
G y can be decomposed into two one dimensional operators as G y = 2 * [− 1 0 1] .
1 

This shows that it performs averaging in vertical direction and differentiation in horizontal
direction.
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

______________________________________________________________________________

QUESTION 5:
A point in cartsian space is represented in Hough space (slope intercept space) as ___________

a. Point

b. Straight line

c. Parabola

d. None of these

Correct Answer: b.

Detailed Solution:

y mx + c .
Equation of line slope-intercept form in Cartesian space is written as=

Let ( m, c ) is Hough space and ( x0 , y 0 ) is a point in Cartesian space, then in Hough space
y 0 = mx0 + c .

1 y
On simplification we write it as m = − c + 0 , which represents a line in ( m, c ) space.
x0 x0

Hence option b is correct.

______________________________________________________________________________
QUESTION 6:

Which of the following statements are true for Laplacian operator?

a) It can be used to find appropriate location of edge using zero crossings.

b) Sensitive to noise.

c) Both a and b

d) None of the above

 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Correct Answer: c.

Detailed Solution:

Lapalcian operator is a second order derivative operator and therefore can be used to find
appropriate location of edge using zero crossings. But this second order derivative makes it
sensitive to noise.

Hence option c is correct.

______________________________________________________________________________

QUESTION 7:
In a given image strip (S), assuming necessary zero padding find second order derivative

S = 2 4 8 10 8 4 2 0 0 0

a) 0 2 2 − 4 − 2 2 0 2 0 0

b) 0 2 − 2 − 4 − 2 2 0 2 0 0

c) 0 2 − 2 − 4 2 2 0 2 0 0

d) 0 2 2 4 2 2 0 2 0 0

Correct Answer: b.

Detailed Solution:
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

Using the relation, f " ( x)= f ( x + 1) + f ( x − 1) − 2 f ( x) . For example for f ( 0 ) ,

=f ( 0 ) f (1) + f ( −1) − 2 f ( 0 ) , here f ( −1) = 0 , therefore f (0) = 4 + 0 − 2 * 2 = 0 .

Similarly, f (1) = f (2) + f (0) − 2 * f (1) == 8 + 2 − 2 * 4 = 2 .

In this way we can compute the second order derivative for this sequence.

Hence option b is correct.

______________________________________________________________________________

QUESTION 8:

Which of the following are major point of interest in discontinuity based segmentation
algorithms?

a. Isolated points
b. Lines
c. Edges
d. All of the above

Correct Answer: d.

Detailed Solution:

In discontinuity segmentation algorithms, isolated points, lines and edges are major point of
interest for segmentation.

Hence option d is correct.

______________________________________________________________________________
 NPTEL Online Certification Courses
Indian Institute of Technology Kharagpur

QUESTION 9:
The applications of segmentation is (are)______________________.

a. Detection of similar regions.

b. Detection of lines and edges in an image.
c. Both a and b.
d. None of the above mentioned.

Correct Answer: c.

Detailed Solution:

Segmentation is used to detect similar regions, lines and edges in an image,

Hence option c is correct.

______________________________________________________________________________

QUESTION 10:

Which of the method is not related to region-based segmentation?

a. Thresholding
b. Merging and splitting
c. Region growing
d. Hough Transform

Correct Answer: d.

Detailed Solution:

Hough transform is not related to region-based segmentation.

Hence option d is correct.
______________________________________________________________________________

************END*******