Faculty Faculty of Science & Technology

Department Department of Computing & Technology

Examination period January 2014

Module Code MOD002643

Module Title Image Processing

Level and credit
Level 6; 15 credits
volume
Duration of
1 hour and 30 minutes
examination
Number of
7
questions set:
Number of pages: 8 (Including Cover)
Name of
Ian van der Linde
examiner/tutor:

Additional materials permitted in the examination?

yes
(in addition to pens/pencils)?
If yes, please complete table below:

Case Study no

Formulae tables no

Graph paper no

Calculator scientific

Other (please specify)

Instructions to Candidates

1. Answer ALL SEVEN questions.

Mitigation – Academic Regulations 6.88 - 6.95

6.94 Claims for mitigation must be submitted by the student or in exceptional circumstances (e.g.
when a student has been hospitalised) by a Director of Studies, Course Group Leader or
Faculty Student Adviser on behalf of the student no later than five working days after the
date on which an examination has been sat.
QUESTIONS

1. The following questions require you to calculate some basic image statistics for the 3×3
4bpp greyscale image L , provided below. Report real numbers to 2 decimal places.

[ ]
7 2 7
L= 5 15 5
7 2 7

(a) State Lmax , , and .

(3 marks)

Lmax −Lmin
C mich=
(b) Calculate the Michelson contrast of this image, given by Lmax +Lmin
(3 marks)

(c) Calculate the RMS contrast of this image, given by

(4 marks)
2. The follow questions require you to perform image thresholding using the 3×3 4bpp
greyscale image , provided below, and two threshold values (T 1=4 ; T 2=12 ).

[ ]
7 2 7
L= 5 15 5
7 2 7

Show the binary matrix produced if we perform the operation L<T 1 , and state the
(a)
name of this operation.
(3 marks)

(b) Show the output matrix if we use the binary matrix you created above to threshold
image .
(3 marks)

(c) Show binary and output matrices produced by , and state the name of
this operation.

(4 marks)
3. The following questions require you to produce histograms for the 3×3 4bpp image ,
provided below. The histograms you produce must be appropriately scaled and axes
must be correctly labelled.

[ ]
8 4 6 5
4 6 9 6
L=
6 9 6 4
5 6 4 8

(a) Show the standard histogram corresponding to image .

(3 marks)

(a) Show the normalized histogram corresponding to image . Report real numbers to 3
decimal places.

(3 marks)

(b) Show the cumulative histogram corresponding to image .

(4 marks)
4. The following questions require you to apply contrast enhancement to a 3×3 4bpp image
, provided below. Round real numbers to the nearest integer, where appropriate.

[ ]
8 4 6 5
4 6 9 6
L=
6 9 6 4
5 6 4 8

( )
max new−min new
Lnew =( L old−minold ) × +min new
max old−minold
(a) Using the formula:

apply the histogram stretch operation to image . Provide both the contrast-
enhanced image matrix, and its corresponding standard histogram.

(10 marks)
5. The following questions require you to perform image rescaling using the 3×3 4bpp
greyscale images , provided below. Round real numbers to the nearest integer, where
appropriate.

[ ] [ ]
10 60 65 45 10 ¿ ¿ 45
55 15 40 70 ¿ ¿ ¿ ¿
L1 = L2 =
50 35 20 75 ¿ ¿ p ¿
30 85 80 25 30 ¿ ¿ 25

(a) Reduce the size of image L1 by a factor of 2 using pixel decimation.

(3 marks)

(b) Reduce the size of image L1 by a factor of 2 using pixel averaging.

(3 marks)

(c) Calculate the value of the missing pixel p in image L2 , using bilinear interpolation.

(4 marks)
6. The following questions require you to perform linear convolution, using the 3×3 4bpp
greyscale image , provided below.

[ ] [ ]
1 1 1 1 2 1
9 9 9 16 16 16

[ ] [ ]
10 20 30 1 0 −1
1 1 1 2 4 2
L= 80 90 40 M 1 = M 2= 2 0 −2 M 3 =
9 9 9 16 16 16
70 60 50 1 0 −1
1 1 1 1 2 1
9 9 9 16 16 16

(a) State the names of spatial domain convolution filters M 1 , M 2 and M 3 .

(3 marks)

(b) Apply M 1 to image (i.e., calculate the matrix produced by L∗M 1 ), leaving the
image border unconvolved. Round real numbers to the nearest integer, where
appropriate.

(3 marks)

(c) Apply M 3 to image (i.e., calculate the matrix produced by L∗M 3 ), leaving the
image border unconvolved, and state how we might use the outcome of this
procedure to sharpen an image. Round real numbers to the nearest integer, where
appropriate.

(4 marks)
7. The following questions require you to understand the spatial frequency domain.

(a) Sketch the spatial domain images that correspond to the following 2D amplitude
spectra.

(i) (ii) (iii)

(3 marks)

(b) Sketch a frequency spectrum mask suitable from removing the periodic noise from
the following image, and briefly state how it would be used.

(3 marks)

(c) State the effect produced by the following spatial frequency domain filters (where
white = pass and black = no pass).

(i) (ii) (iii) (iv)

(4 marks)