Low-Light Image Enhancement

This project focuses on enhancing low-light images using various preprocessing, enhancement, and postprocessing techniques. The system automatically detects low-light conditions and applies a series of adjustments to improve image quality.

Features

• Automatic Low-Light Detection: Determines if an image requires enhancement.
• Noise Reduction: Removes noise while preserving image details.
• Brightness and Contrast Adjustment: Optimizes brightness and contrast for better visibility.
• Detail Enhancement: Sharpens and highlights important features in the image.
• Color Balance Adjustment: Applies white balance correction for natural-looking colors.
• Postprocessing: Uses guided filtering for final touch-ups.

Dependencies

Ensure you have the following Python packages installed:

• opencv-python
• numpy
• scikit-image
• matplotlib
• pytest

You can install these dependencies using pip:

pip install opencv-python-headless numpy scikit-image matplotlib pytest

Project Structure

low-light-image-enhancement/
├── src/
│   ├── main.py              # Main script to process images
│   ├── enhancement/
│   │   ├── core.py          # Core enhancement functions
│   │   ├── preprocess.py    # Preprocessing utilities
│   │   └── postprocess.py   # Postprocessing utilities
├── test/
│   ├── test_enhancement.py  # Tests for enhancement functions
│   ├── test_utils.py        # Tests for utility functions
├── data/
│   ├── input/               # Directory for input images
│   ├── output/              # Directory for output images
│   ├── psnr/                # Directory for test images
├── README.md                # Project documentation
└── requirements.txt         # List of dependencies

Usage

1. Prepare Input Images: Place low-light images in the data/input directory.
2. Run the Script: Execute the following command:

   python src/main.py

3. Check Output: Enhanced images will be saved in the data/output directory.

How It Works

1. Low-Light Detection:

   • The system checks the mean intensity of the grayscale version of the image.
   • If the mean intensity is below a threshold (default: 50), the image is considered low light.

2. Enhancement Pipeline:

   • Preprocessing: Noise reduction using bilateral filtering.
   • Brightness & Contrast Adjustment: Gamma correction to enhance brightness.
   • Detail Enhancement: Unsharp masking for sharpening.
   • Color Balance Adjustment: White balance using LAB color space.
   • Postprocessing: Guided filtering for smooth and visually pleasing results.

Metrics for Evaluation

• PSNR (Peak Signal-to-Noise Ratio): Measures signal fidelity.
• SSIM (Structural Similarity Index): Evaluates structural similarity between original and enhanced images.

Example

Input Image	Enhanced Image

Customization

You can modify the enhancement pipeline in the core.py file. For instance:

• Adjust gamma values for brightness correction.
• Modify filtering parameters for noise reduction.
• Change thresholds for low-light detection.