Image Convolution with Custom Kernel

This project applies a convolution operation to a grayscale image using a custom kernel. The program demonstrates the core concepts of kernel convolution, which is commonly used in computer vision and convolutional neural networks (CNNs) for image filtering and feature extraction.

Project Structure

├── images/                 # Directory for input images
│   ├── dog_picture.jpg     # Example image used for convolution
├── lernel_convolution.py   # Python script implementing the convolution operation
├── README.md               # Project documentation
└── requirements.txt        # Python dependencies

Features

• Convolution Operation: Applies a 5x5 Gaussian blur filter using a custom kernel.
• Image Padding: Pads the original image with zeros to handle edge pixels during convolution.
• Pixel Processing: Each pixel in the image is processed using the convolution operation, and a new filtered image is generated.
• Output: The filtered image is displayed after applying the convolution.

Requirements

• Python 3.8+
• Pillow (PIL)

You can install the dependencies by running:

pip install -r requirements.txt

Usage

1. Place your input image in the images/ directory.
2. Modify the kernel_convolution.py script to load your image file.
3. Run the script to apply the convolution:

   python kernel_convolution.py

   The script reads the input image, converts it to grayscale, applies a 5x5 Gaussian blur kernel, and displays the resulting filtered image.

Learning Objectives

This project was used to understand:

• How convolution works in image processing.
• The role of kernels in feature extraction and image filtering.
• The fundamental concepts behind convolutional neural networks (CNNs).

Future Work

• Experiment with different kernels for edge detection, sharpening, and other image transformations.
• Extend the project to process images in batches or integrate it into a more complex CNN pipeline.