OpenCV Image Processing Application

A PyQt5-based application for demonstrating image processing techniques using OpenCV.

（成大資工系｜影像處理、電腦視覺及深度學習概論｜作業）

Features

1. Image Processing

• 1.1. Color Separation
• 1.2. Color Transformation
• 1.3. Color Detection
• 1.4. Blending

2. Image Smoothing

• 2.1. Gaussian Blur
• 2.2. Bilateral Filter
• 2.3. Median Filter

3. Edge Detection

• 3.1. Gaussian Blur
• 3.2. Sobel X (Vertical Edge Detection)
• 3.3. Sobel Y (Horizontal Edge Detection)
• 3.4. Magnitude

4. Transforms

• 4.1. Resize
• 4.2. Translation
• 4.3. Rotation & Scaling
• 4.4. Shearing

Requirements

Python Environment

• Python 3.7

Core Dependencies

• opencv-contrib-python (3.4.2.17)
• Matplotlib 3.1.1
• PyQt5 (5.15.1)
• numpy (>=1.19.0)

Additional Dependencies

• numpy
• pillow

Required Images

Image Processing

• OpenCV.png (for color operations)
• Dog_Strong.jpg and Dog_Weak.jpg (for blending)
• beauty.png (for smoothing operations)
• building.jpg (for edge detection)
• Microsoft.png (for transforms)

Installation and Setup

General Setup

1. Clone the repository:

git clone [your-repository-url]
cd [repository-name]

2. Create virtual environment:

python -m venv venv

Windows Setup

venv\Scripts\activate
pip install -r requirements.txt

Linux/Mac Setup

source venv/bin/activate
pip install -r requirements.txt

Usage

Image Processing Operations

Basic Operation

1. Run the application:

python main.py

2. Using the Interface:
   • Use "Load Image 1" for single image operations
   • Use both "Load Image 1" and "Load Image 2" for blending
   • Click operation buttons to perform specific tasks
   • Use trackbars in popup windows to adjust parameters
   • Press 'Esc' to close popup windows

Feature-Specific Operations

Edge Detection

1. Load building.jpg using "Load Image 1" button
2. Select operations in sequence:
   • "3.1 Gaussian Blur" for initial smoothing
   • "3.2 Sobel X" for vertical edges
   • "3.3 Sobel Y" for horizontal edges
   • "3.4 Magnitude" for combined edge strength

Transform Operations

1. Load Microsoft.png using "Load Image 1" button
2. Apply transformations in sequence:
   • "4.1 Resize" (430x430 → 215x215, centered at (108, 108))
   • "4.2 Translation" (shift from (108, 108) to (323, 323))
   • "4.3 Rotation, Scaling" (45° counter-clockwise, 0.5x scale)
   • "4.4 Shearing" (points transform: [[50,50],[200,50],[50,200]] → [[10,100],[100,50],[100,250]])

Technical Details

Project Structure

project/
│
├── main.py           # Main application entry point
├── controller.py     # Main logic and image processing functions
├── UI.py            # PyQt5 UI definition
│
├── images/          # Sample images directory
│   ├── OpenCV.png
│   ├── Dog_Strong.jpg
│   ├── Dog_Weak.jpg
│   ├── beauty.png
│   ├── building.jpg
│   └── Microsoft.png
│
└── requirements.txt  # Project dependencies

Implementation Details

Edge Detection Kernels

• Gaussian Kernel (3x3):

[[0.045 0.122 0.045]
 [0.122 0.332 0.122]
 [0.045 0.122 0.045]]

• Sobel X Kernel:

[[-1  0  1]
 [-2  0  2]
 [-1  0  1]]

• Sobel Y Kernel:

[[ 1  2  1]
 [ 0  0  0]
 [-1 -2 -1]]

Edge Magnitude Formula

Magnitude = sqrt(SobelX² + SobelY²)
Normalized to range [0, 255]

Notes

• All image paths should be relative to the project directory
• Application uses OpenCV's default BGR color space
• Smoothing operations use odd-numbered kernel sizes (2m+1)
• Custom implementations avoid built-in OpenCV functions for edge detection

Controls

• ESC: Close popup windows
• Trackbars: Adjust parameters for image processing operations

Troubleshooting

• If image windows don't close properly, press any key and then 'ESC'
• For optimal performance, ensure input images match the recommended dimensions
• If trackbars aren't responding, click on the image window to ensure it's in focus
• Common dimension issues:
  • Edge detection expects images of sufficient resolution for clear edge detection
  • Transform operations are optimized for 430x430 input images
  • Resizing operations maintain aspect ratio by default

Reference Formula

Gaussian Blur

G(x,y) = (1/(2πσ²)) * e^(-(x²+y²)/(2σ²))
where σ = √0.5

Color Transformation

Grayscale conversion:
I1 = 0.07*B + 0.72*G + 0.21*R  (OpenCV function)
I2 = (R+G+B)/3                  (Average weighted)

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is available for academic and educational purposes.

Acknowledgments

• Based on OpenCV computer vision library
• Created as part of image processing coursework