Lincoln School of Computer Science
Assessment Component Briefing Document
Title: CMP3108M Image Processing, Assessment One
Indicative Weighting: 100%
Learning Outcomes:
On successful completion of this component the student will have demonstrated competence in the following areas:
• [LO1] critique the theoretical knowledge of image processing, including how to process and extract quantifiable information from images.
• [LO2] apply image processing techniques to solve practical problems.
Requirements
This assessment comprises two assessed components, as detailed in the following page.
1. A report (in PDF format) that describes your approach to the tasks (maximum 4 pages, including figures but not the cover page). Weighting: 40% of this assessment.
2. A file containing all functions written in the MATLAB code with clear comments and requested figures. Weighting: 60% of this assessment.
Tasks
Download and unzip the file ‘Assignment_Input.zip’ from Blackboard. You should obtain:
• A dataset of 16 images containing a white swan logo
• Two MATLAB script m-files named ‘Task1to3.m’ to ‘Task4.m’.
Complete the MATLAB m-files to perform the corresponding tasks described below. As a guide, a few command lines for performing the Tasks have already been added to the script. You need to add the command lines to implement the other steps. Ensure you add appropriate comments to your code to briefly explain what each section is doing. You CAN use any built-in function but not any custom functions written by others (e.g. from Matlab File Exchange).
You must make an electronic submission of your report in PDF format. The electronic submission should also include the two MATLAB m-files, as supporting materials, which produce the desired results and display the outputs. Make sure the MATLAB scripts are correct and functional and do not display any error message. Put all the files (excluding the provided images and compress the folder into a zip file for submission. Name your zip file and PDF report using this format: LastName_ FirstName_StudentNo.
Task1 – Pre-processing (20%)
Add code to the MATLAB script Task1to3.m to load the image ‘IMG_01.jpg’ and convert it to grey- scale. Then reduce the image size from its original size to half by bilinear interpolation. With the re-
sized image, produce its histogram. Based on the histogram, discuss and justify in your report what value would be a good threshold for binarising the image so that the swan logo can be detected. Then generate the binarised image. Display the re-sized image, histogram and binary image in your report.
Task2 – Edge Detection (20%)
Continue to write your code in the MATLAB script Task1to3.m to apply edge detection techniques (both Sobel and Canny) to the grey-scale image you produced in Task 1. Display the detected results and add one close-up (zoomed-in section) to you report where the difference between the two edge sections of the swan is clear. Discuss the differences you notice between the two techniques.
Task3 – Simple Swan Segmentation (20%)
Complete the MATLAB script Task1to3.m to automatically segment the swan using either the binary image obtained in Task1 or edge image obtained from Task 2. Write down the steps you took in the report and include the resulted image(s). Note the segmented image should be binary and only contain the complete white swan (as accurate as possible) on a black background. (Hint: use connected component analysis)
Task4 – Swan Recognition (40%)
You will likely to find that the script you’ve written so far will not work if you change ‘IMG_01.jpg’ to another image from the provided dataset. Complete the MATLAB script Task4.m to recognise and segment the swan through a series of image processing techniques you choose, generate a binary image where zero means no swan detected and a non-zero value means that the pixel belongs to the swan logo as shown in the figure below. For this task, your target is to write an automatic and robust method that is able to accurately detect the swan logo from all provided images.
Tip: One way to carry out such task is to apply colour thresholding; carry out connected component analysis to extract regions of interest; then use region and shape features to recognise the segments that belongs to a swan.
Your solution must minimise the amount of hard thresholds in order to make the algorithm as robust as you can. It should NOT involve any training (i.e. a machine learning based approach).
In your report, explain each step you have taken and why you have used it. Select one particular image other than ‘IMG_01.jpg’, and illustrate the outcome of each processing stage by adding example figure(s) to your report. Finally, evaluate your method by reporting its performance on the entire dataset, e.g. state for which images your solution worked and which ones it didn’t work.
For Task4, you can optionally save all the resulted images when running the algorithm on the entire dataset in a file called ‘output’. This output file can be zipped up together with your code and submitted through ‘supporting material upload’.
Useful Information
This assessment is an individually assessed component. Your work must be presented according to the Lincoln School of Computer Science guidelines for the presentation of assessed written work. Please make sure you have a clear understanding of the grading principles for this component as detailed in the accompanying Criterion Reference Grid.
If you are unsure about any aspect of this assessment component, please seek the advice of a member
of the delivery team.
Submission Instructions
The deadline for submission of this work is included in the School Submission dates on Blackboard.
You must make an electronic submission of your report in PDF format together with a zip file containing all source code files by using the assessment link on Blackboard for this component. The report should be submitted through TurnItIn and the zip file should be uploaded as supporting material. You must attend the lectures for further details, guidance and clarifications regarding these instructions.
DO NOT include this briefing document with your submission.
Reviews
There are no reviews yet.