reissvd_imgcompress

A Python package for image compression using the REIS-SVD algorithm. This package helps in reducing the size of images while preserving quality, and also provides various metrics for evaluating compression quality, including MSE, PSNR, and SSIM.

Update

• Remove the limitation on k: The code allows k to exceed the total number of singular values and sets k to the maximum available value if necessary to avoid errors.
• Improve the precision of the Sylvester equation: Increase max_iter to 50 iterations and reduce tol to 1e-6, which allows for a more accurate solution of the equation.

Features

• Compress grayscale images using REIS-SVD.
• Compute compression metrics:
  • Compression Ratio (CR)
  • Mean Squared Error (MSE)
  • Peak Signal-to-Noise Ratio (PSNR)
  • Structural Similarity Index (SSIM)
• Visualize compressed images with different singular values.

Installation

To install the package, first clone the repository or download the code, then navigate to the package directory and run:

pip install .

Usage

from reissvd_imgcompress import (
    rgb2gray, reis_svd, compute_metrics, plot_compressed_images
)
import numpy as np
from PIL import Image

# Load an image and convert it to grayscale
image = np.array(Image.open('sample_image.png'))
gray_image = rgb2gray(image) # If grayscale conversion is not needed, you can skip this step.


# Define singular value counts for testing
k_values = [1, 2, 5, 10, 20, 50, 100, 200, 500]

# Iterate over k_values, compress images, and compute metrics
for k in k_values:
    compressed = reis_svd(gray_image, k)
    CR, mse, psnr, ssim_index = compute_metrics(gray_image, compressed, k)
    print(f"k={k}: CR={CR:.2f}, MSE={mse:.2f}, PSNR={psnr:.2f} dB, SSIM={ssim_index:.4f}")

# Plot compressed images for visual comparison
plot_compressed_images(gray_image, k_values)