MARKING CRITERIA
There are two questions in the assignment.
Question 1:
Topic Sub-topic Details
Remove the periodic noise FFT or DFT
transform Implementation + coding
Showing results in the frequency domain with the noise
Design frequency domain filters to remove the periodic noise Design + implementation + coding +comparison
Showing the results and discussion
Remove random noise Algorithm design + implementation + coding Design various filters, eg. low-pass, median filters…, and
comparison
Display results and discussion
Calculate the error + discussion
Question 2
It is a research exercise. Students are expected to carry on literature review and to define measures for image quality.
Topic
Literature review on image quality evaluation
Define a measure of image quality for the case presented in this question and justify the definition
Design a de-blur algorithm
Results and discussion
ASSIGNMENT DETAILS
Coursework assignment 1 – image enhancement
(a) Original image (b) Distorted image Figure 1. Dog images
1. The original grey level image (dogOriginal.bmp) and distorted image (dogNoise.bmp) are given in Figures 1(a) and 1(b), respectively. It is known that the distortion is caused by combination of random noise and periodic noise. You are required
• to develop algorithms in both the spatial domain and the frequency domain to improve the quality of the distorted image by removing the noises; and
• to evaluate your results by comparison of the restored images (after applying your algorithms to the distorted image) with the original image by estimation of the average mean square error.
(You may present findings from the exercise in the report.)
2. The image shown in Figure 2 was blurred by unknown reasons (such as hand-shaking during capturing, PSF in the optical system…). You are asked to improve the image quality by using image enhancement techniques, and show that image qualities are improved.
• Devise algorithms which work for the purpose;
• Research on image quality evaluation methodology;
• Based on the research outcome, define your own measures which quantitatively evaluate image quality improvement.
(Make assumptions where needed.)
Figure 2. Poor quality images – blurred wood
A formal report is required (maximum 8 A4 pages excluding preamble and appendices). It should include (with descriptive material) all the details of the processes that show how you carried out the tasks. You may implement your algorithms in Matlab, The original code (with detailed comments) should be attached at the end of the report. Limited Matlab/OpenCV functions may be used in the implementation with underlying techniques explained clearly.