EGH444: Digital Signals and Image Processing QUT - Semester 2, 2020
Assessed Lab Task 1
Due: Friday 4 September 2020 (Week 7)
Unit Coordinator & Lecturer: A/Prof. Thierry Peynot
Tutors: Osman Tursun & Jordan Laurie
Notes
• Sample images mentioned by name in the text should be downloaded from Blackboard.
• Note this is an individual assessment, each student should submit their own answer.
• All your answers should be entered in the single template file: EGH444 Task1 Surname StudentNb.m,
renamed to include your surname and student number.
• In this file, answers should be provided below the corresponding question as indicated.
• Text answers, wherever requested, should be commented out1.
• Code scripts should have no dependencies (other than specified input images where required),
and no file outputs. (i.e. figures/plots/workspace only)
• The code should run on Matlab R2019b (version of Matlab in the computer labs) without error.
• The total number of marks for this tasks is 10 (10% of the unit). The maximum number of marks
for each question is indicated in parenthesis (x pts).
1However, if you think some illustration may help to clarify your answer, you may add some (short) section code to
illustrate your point, for example generating a relevant figure.
1
Question 1 - Theory (1pt)
1. Suppose a digital image, A, is subjected to histogram equalisation to create image A1. What
distribution best describes the histogram of pixel intensities in A1? (Text) (0.5pt)
2. If A1 is equalised in the same manner to create A2, is the distribution of pixel intensities the
same, or different, to that of A1? Why? (Text) (0.5pt)
Question 2 - Code (3pts)
Write the appropriate Matlab code to manually stretch the dynamic range of a given 8-bit grayscale
image such that:
• values in the range [0..80] become 0,
• those in the range [81..120] are rescaled into the range [0..120],
• grey-levels [120..150] are rescaled in the range [151..255],
• and higher values are all set to 255.
Display the resulting image in Figure 1 and its histogram in Figure 2.
Display the original image after applying histogram equalisation in Figure 3.
Compare the two images obtained and comment on the differences.
Also compare the two histograms and comment on the differences.
NB: You may test the code on the image moon.tif but note the marker will also be testing it with
another test image.
Question 3 - Conceptual Understanding/ Professional Development
(1pt)
Why is interpolation needed when resizing an image? (Text)
Question 4 - Application and Code (3pts)
Consider a 3×3 spatial mask that averages all neighbours of a point (x, y) in this 3×3 neighbourhood,
but excluding the point itself.
1. Find the expression of the equivalent filter, H(u, v), in the frequency domain. (Provide the
equation H(u, v)). (Text) (1pt).
(Hint: it may help to start by writing the filtered image as: g(x, y) = . . . , function of the original
image f .)
2. Code a function that applies this filter to an image and displays the original image and the filtered
image. (Code) (1pt)
3. Briefly comment on the appearance of the filtered image. (Text) (1pt)
2
Question 5 - Application and Code (2pts)
1. The image ‘NASA appolo17 noisy.tif’ (courtesy of NASA) is corrupted by sinusoidal noise.
Write the code to execute a Butterworth notch filter in the frequency domain on this image that
eliminates this noise (such as the one represented in ‘BW bandreject order4.tif’). (Code)
(1pt)
2. Display the original image, the filtered image and the image of difference between these two.
Briefly comment on the results obtained. (Code and Text) (1pt)
3