Description
Problem: Investigate the effectiveness of different image filtering and thresholding
algorithms on different images
In this assignment, the investigation will be conducted in 2 stages:
Stage 1: Image filtering and thresholding (7 marks)
Objective: Compare and analyse the effects of filtering (median and Gaussian filtering) and
thresholding (global, Otsu and adaptive thresholding) algorithms
Two images to be experimented on:
image1.png image2.png
To do:
1. Test the performance of global, Otsu and adaptive thresholding on the original
images
a. Which thresholding algorithm performs the best for each image?
2. Apply median filtering on the original images, and then test the performance of
global, Otsu and adaptive thresholding on the filtered images
a. Which thresholding algorithm performs the best for each image?
b. What parameters are optimal for the median filtering?
3. Apply Gaussian filtering on the original images, and then test the performance of
global, Otsu and adaptive thresholding on the filtered images
a. Which thresholding algorithm performs the best for each image?
b. What parameters are optimal for the Gaussian filtering?
Present the visual results and discussion your findings to the questions in the report. Give
some brief theoretical analysis about the different effects of filtering and thresholding
algorithms on the two images, i.e. why do these algorithms achieve different results on the
two images?
Stage 2: Use filtering and thresholding for segmentation (3 marks)
Objective: Segment (i.e. separate) the cells (dark and light gray round-like objects) from the
background using simple image filtering and thresholding
One image (image3.jpg) to be experimented on:
To do: Find a combination of the filtering (median, Gaussian) and thresholding (global, Otsu
and adaptive) steps that can achieve a good segmentation result
Present the visual segmentation results and describe how the segmentation was performed
in the report. Pseudo code is acceptable as the description. Need to include the choice of
parameters. Give a brief discussion about why the designed method could work.
Note: a visual segmentation result is an image of the same size as the input image, with
pixels of the segmented cells shown as black and the background pixels shown as white.
REFERENCES
[1].https://opencv-pythontutroals.readthedocs.io/en/latest/py_tutorials/py_tutorials.html