Nptel Online Certification Course

Indian Institute of Technology Kharagpur

Computer Vision
Assignment - Week 6
Number of questions: 10 Total marks: 10x2=20

QUESTION 1: Type: MSQ

Which of the following is/are true for SIFT and SURF?

a) Computation of scale space and histograms of gradient direction are approximated in

SURF compared to SIFT.

b) Integral of images are used in SURF to speed up computation.

c) Small Gaussians used in SIFT are approximated by box filters in SURF.

d) Haar wavelet responses are used to accumulate the orientation in SIFT and SURF
descriptors.

Correct Answer: a, b, c
Detailed Solution:

1
QUESTION 2: Type: MSQ
Which of the following is/are true?

a) RANSAC is best suited for fitting outliers..

b) Hough transform is best suited for line fitting in presence of many lines.

c) Kullback-Leiber Divergence is best suited to find the ’optimal’ line parameters when
points belonging to a line are given.

d) The average case time complexity of NN Search using K-D tree is 𝑂 (𝑙𝑜𝑔(𝑁) + 2𝑑 ) where
𝑑 is the dimension of the feature space.

Correct Answer: a, b, d
Detailed Solution:

2
QUESTION 3: Type: MCQ
3 2 2
Consider the set of parametric equations 𝑥 = 𝑡 + 2𝑡 and 𝑦 = 𝑡 + 𝑡. representing a 2D curve.
Compute the tangent vector at 𝑡 = 2.

a) (16, 6)

b) (16, 5)

c) (20, 5)

d) (20, 6)

Correct Answer: c)
Detailed Solution:
Perform derivation 𝑥 ′ = 3𝑡 2 + 4𝑡 and 𝑦′ = 2𝑡 + 1. Compute the tangent at 𝑡 = 2.

3
QUESTION 4: Type: MCQ
Consider the following matrix representing 4 sample data points (1, 2), (2, 1), (3, 2), (2, 3) as
1 2
 
2 1
𝑋=   . Compute the covariance matrix using data points from 𝑋.
3 2 

2 3
 
 
1 0
a)
0 1
 
0.5 0
b)
0 0.5
 
0.25 0
c)
0 0.25
 
2 0
d)
0 2

Correct Answer: b)
Detailed Solution:
Compute the mean vectors as [ 1+2+3+2
4 , 2+1+2+3
4 ] = [2, 2]. Now compute
1 − 2 2 − 2 −1 0 
   
2 − 2 1 − 2  0 −1
𝑋˜ = 

 =  1 0 .
  
3 − 2 2 − 2   
2 − 2 3 − 2  0 1 
   
˜𝑇 ˜
The covariance matrix is given by 𝑋 4 𝑋 .

4
QUESTION 5: Type: MCQ
The top left corner of the matrix is considered as (1, 1) and indices increment for 𝑥 and 𝑦 axes
along horizontal right and vertical downward directions, respectively.
0.3 0.1 0.6
 
Consider a 3 × 3 matrix 𝐴 = 0.5 0.2 0.3 , compute the magnitude and orientation (in
0.1 0.3 0.4
 
degrees) using first order first difference operator at the center pixel, respectively.

a) 0.2828 and −90

b) 0.1414 and −45

c) 0.2828 and 90

d) 0.1414 and 45

Correct Answer: b
Detailed Solution: √
( 𝑓 (𝑥 𝑖+1 ,𝑦 𝑖 )− 𝑓 (𝑥 𝑖−1 ,𝑦 𝑖 )) 2 +( 𝑓 (𝑥 𝑖 ,𝑦 𝑖+1 )− 𝑓 (𝑥 𝑖 ,𝑦 𝑖−1 )) 2
Magnitude at a pixel location (𝑥𝑖 , 𝑦𝑖 ) is 2 and
−1
orientation is 𝑡𝑎𝑛 ( 𝑓 (𝑥𝑖 , 𝑦𝑖+1 ) − 𝑓 (𝑥𝑖 , 𝑦𝑖−1 ))/( 𝑓 (𝑥𝑖+1 , 𝑦𝑖 ) − 𝑓 (𝑥𝑖−1 , 𝑦𝑖 )), where the function
𝑓 () gives out the intensity at that particular location.

5
QUESTION 6: Type: MCQ
If an octave in SIFT operator has 10 images, then the scale factor by which each image in
octave differs from the next image is

a) 1.080

b) 1.065

c) 1.071

d) 1.090

Correct Answer: a)
Detailed Solution:
Within an octave, the adjacent scales differ by a constant factor 𝑘. If an octave contains 𝑠 + 1
images, then 𝑘 = 21/𝑠 . The scale factor is 2 in an octave. Therefore, 𝑘 = 21/9 .

6
QUESTIONS 7: Type: MCQ
Given two features each of size 7 as 𝑓 (2, 4, 1, 3, 1, 1, 7) and 𝑔(5, 3, 5, 2, 1, 1, 8). Compute 𝐿 1
and 𝐿 2 norms between 𝑓 and 𝑔, respectively.

a) 4 and 3.16

b) 4 and 2

c) 10 and 5.29

d) 10 and 3.16

Correct Answer: c)
Detailed Solution:
| 𝑓 − 𝑔| 2 ) 1/2 .
Í𝑛 Í𝑛
𝐿 1 norm is 𝑖=0 | 𝑓 − 𝑔| and 𝐿 2 norm is ( 𝑖=0

7
QUESTIONS 8: Type: MSQ
Which of the following is/are true for edge and corner detection?

a) A corner will have low maximal and minimal curvatures.

b) A corner will have high maximal and minimal curvatures.

c) An edge will have low maximal curvature, but very high minimal curvature.

d) An edge will have high maximal curvature, but very low minimal curvature.

Correct Answer: b), d)

Detailed Solution:
A corner will have high maximal and minimal curvatures. An edge will have high maximal
curvature, but very low minimal curvature.

8
QUESTION 9: Type: Numeric
Given two features each of size 5 as 𝑓 (3, 2, 4, 5, 1) and 𝑔(7, 2, 5, 3, 2). Compute 𝐿 1 norm
between 𝑓 and 𝑔.
Correct Answer: 8
Detailed Solution:
Í𝑛
𝐿 1 norm is 𝑖=0 | 𝑓 − 𝑔|.

9
QUESTION 10:   Type:MCQ
1 2
Find the eigen values of the matrix .
2 1

a) −1, 2

b) 1, −2

c) −1, 3

d) 1, −3

Correct Answer: c)
Detailed Solution:
1−𝜆 2
The eigen values are calculates as = 0.
2 1−𝜆

10