- Bruno Flávio Ferreira
- Edson Yudi Toma
The project aims to compress an image using Fourier Transform and Huffman Compression.
The input images can be any uncompressed image, we use in this project the "Standard" test images, an .tif image database located at http://imageprocessingplace.com/root_files_V3/image_databases.htm to test the algorithm.
The images can be colored or not:
- Read uncompressed image
- Apply FFT in the image
- Cut of values below an given threashold
- Apply Huffman compression in the FFT image
- Save the compressed image
OBS: The threshold value is a percentage of maximum value
- Read compressed image
- Apply Huffman decompression
- Aplly the Inverse FFT in the decompressed image
- Show the image in the screen
| Image | Original Size | Threshold = 0.008 | Threshold = 0.01 | Threshold = 0.05 |
|---|---|---|---|---|
| cameraman.tif | 262,8 kB | 245,0 kB | 217,1 kB | 138,3 kB |
| lena_color_512.tif | 787,5 kB | 478,3 kB | 454,9 kB | 399,1 kB |
| mandril_color.tif | 787,4 kB | 661,6 kB | 553,6 kB | 399,6 kB |
The program was able to compress the image, but with a loss of quality. This is due to the fact that there are few values repeated after the FFT application, together with the need to save the huffman tree, it is necessary to cut many values (ie a high threshold) to decress the image size, resulting in loss of quality.
| Original Image | Threshold = 0.008 | Threshold = 0.01 | Threshold = 0.05 |
|---|---|---|---|
 |
 |
 |
 |
It is possible to observe that with very high threshold value, the resulting image is incomprehensible.
To run this program you will need Python3 installed in your machine and some libraries:
- Numpy
- Scipy
- Imageio
- Pillow
To install then you can run the following command:
pip3 install numpy
pip3 install scipy
pip3 install imageio
pip3 install pillow
You may have to install python-tkinter for the program to work correctly, on Ubuntu you can execute this command:
sudo apt-get install python3-tk
To run this program simply execute the run.sh, passing the desired arguments.
To view the possible arguments, execute: ./run.sh -h
To compress an image, run:
./run.sh -c path-to-image -o output-image
This will compress the image with the default threshold of 0.01, to especify a threshold, run:
./run.sh -c path-to-image -o output-image -t value
To decompress an image, run:
./run.sh -d path-to-image
To test this program you can run the following command:
./run.sh -c standard_test_images/mandril_color.tif -o mandril.bin
./run.sh -d mandril.bin
This will compress the mandril image, decompress and show on screen