🎨 Image Segmentation using Graph-Based Region Merging

Efficient and adaptive image segmentation with Kruskal-based graph merging and interactive WinForms UI.

---

🛠️ Setup Instructions

✅ Built with C# (.NET Framework 4.8)
🖼️ UI: Windows Forms (WinForms)
💻 OS: Windows 10 or later

⚙️ To run:

1. Open the .sln file using Visual Studio 2019+.
2. Make sure target framework is set to .NET Framework 4.8.
3. Run the MainForm.cs.
4. Load any image (.jpg, .png, .bmp) and start segmenting.

---

📷 Screenshots

---

🧠 Project Overview

This project implements graph-based image segmentation based on the method by Felzenszwalb and Huttenlocher, providing:

• Accurate and scalable segmentation
• Visualization of distinct regions via coloring
• User interaction to highlight or extract segments
• Output of segmentation statistics

It segments color images by analyzing R/G/B channels separately and intersecting the results.

---

✨ Features

• ✅ Gaussian filter smoothing (1D) with configurable σ and mask size
• 📊 Graph construction with 8-connected pixel neighbors
• 🧠 Intelligent region merging based on internal/external difference
• 🖌️ Segment coloring with optional manual selection
• 📄 Output file with total components and sorted sizes
• ⚡ Parallel processing across R/G/B channels
• 🔍 High-resolution image support (tested up to 6720×4480)
• 🪢Choose Segments From the image

---

📈 Algorithm Details

🧮 Step-by-step pipeline

1. Gaussian Smoothing: Reduces noise using a separable 1D Gaussian filter.
2. Graph Construction:
   • Each pixel = node
   • Connect to 8 neighbors
   • Edge weight = |I(pi) - I(pj)| (per channel)
3. Segmentation per Channel:
   • Sort edges by weight
   • Merge with DSU using internal difference + adaptive threshold
4. Color Image Merging:
   • Segment R, G, B channels separately
   • Intersect segment labels for final result
5. Colorize Output: Assign a unique color to each segment
6. Interactive UI:
   • Click to select segments
   • Highlight and export selected area

---

🔬 Why Kruskal?

Unlike basic region-growing, this method:

• Is global (looks at the entire image graph)
• Uses adaptive merging (based on size & internal variability)
• Produces meaningful regions — not just locally similar pixels

📚 Based on:

P. Felzenszwalb and D. Huttenlocher,
Efficient Graph-Based Image Segmentation,
Int. Journal of Computer Vision, 2004.

---

🗂️ Project Structure

ImageSegmentation/
│
├── Segmenter.cs             # Core segmentation logic
├── DisjointSet.cs           # DSU with path compression
├── GraphBuilder.cs          # Builds 8-connected graph
├── ImageOperations.cs       # Image load + Gaussian blur
├── MainForm.cs              # WinForms UI  
├── output.txt               # Segment sizes
├── screenshots/             # Sample output images

---

💻 How to Use

1. Open the program
2. Click “Open” → Choose an image
3. Adjust Gaussian Filter settings if needed
4. Click “Apply Segmentation”
5. Click Merge Segments to choose specific segments
6. Use mouse to click on regions to highlight
7. Click Save to save the segmented image

---