This project demonstrates image compression using Discrete Wavelet Transform (DWT) with configurable hard thresholding. The method reduces the image file size while attempting to preserve its visual quality.
Wavelet compression is a powerful technique that transforms an image into its frequency components. By thresholding and removing insignificant coefficients, we can reconstruct a near-identical version of the image with much smaller size.
This implementation uses:
pywt(PyWavelets) for DWTOpenCVandmatplotlibfor image I/O and visualizationNumPyfor matrix operations
- Compresses grayscale images using 2D wavelet transform
- Adjustable compression threshold
- Visualizes the original vs. compressed images
- Calculates and prints compression ratio and space savings
- Supports batch testing on multiple thresholds
Install the required packages via pip:
pip install numpy matplotlib opencv-python PyWavelets▶ How to Run
- Place your grayscale image (e.g. lena30.jpg, 4K.jpg) in the same directory.
- Run the Python script:
python wavelet_compression.py
- You’ll see original and compressed images side by side, along with the compression statistics printed to the terminal.
Example Output
[Threshold=1] Original: 200.54 KB | Compressed: 190.12 KB | Ratio: 1.05 [Threshold=32] Original: 200.54 KB | Compressed: 83.45 KB | Ratio: 2.40
And visual comparison:
Original Threshold = 8 Threshold = 32
How It Works?
- Wavelet Decomposition: The image is decomposed into approximation and detail coefficients using DWT.
- Thresholding: Coefficients with absolute values below a set threshold are zeroed.
- Reconstruction: The image is reconstructed from the modified coefficients.
- Compression Ratio: File sizes before and after compression are compared.
Parameters to Tune • wavelet type: 'haar', 'db1', 'db2', etc. • level: Decomposition level (default is 2) • threshold: The cutoff for discarding coefficients
Author
Developed by Amir-Abbas Alvand – feel free to contribute or reach out for collaboration!