🚀 VisionSeg: Enhanced Image Segmentation Framework

📌 Overview

VisionSeg is a modular computer vision framework for classical image segmentation, designed with software engineering principles.

Unlike simple scripts, this project is built as a scalable system with:

• Config-driven pipelines
• Modular architecture
• Extendable components

It focuses on separating foreground objects from background using traditional computer vision techniques.

---

🎯 Key Features

🔹 Segmentation Methods

• Fixed Thresholding
• Otsu Thresholding
• Adaptive Thresholding

🔹 Morphological Processing

• Erosion
• Dilation
• Opening (noise removal)
• Closing (hole filling)

🔹 Object Analysis

• Connected Components
• Object Counting
• Area Filtering

🔹 Visualization

• Bounding Boxes
• Object Highlighting
• Result Export

🔹 System Features

• Config-driven pipeline (config.yaml)
• Modular code structure
• Easy extensibility

---

🧠 Concepts Covered

• Image Segmentation
• Binary Thresholding
• Morphological Transformations
• Noise Reduction
• Connected Component Analysis
• Pipeline Design in Computer Vision

---

🏗️ Flow of the Project & Technologies

Technologies Used:

• Python: Core programming language.
• OpenCV (cv2): Image processing, thresholding, morphology, and component analysis.
• NumPy: Matrix operations and statistical calculations (e.g., median area).
• PyYAML: Configuration management.

Pipeline Flow:

1. Input Image
     ↓
2. Preprocessing (Grayscale, CLAHE Contrast Enhancement, Gaussian Blur)
     ↓
3. Segmentation (Otsu / Adaptive / Fixed Thresholding)
     ↓
4. Morphology (Opening, Closing, Dilating, Eroding for Noise Removal)
     ↓
5. Component Analysis (Connected Components, Area Filtering, Circularity)
     ↓
6. Visualization & Output (Bounding Boxes, Highlighted Objects)

---

📁 Project Structure

visionseg/
│
├── main.py
├── config.yaml
├── requirements.txt
├── README.md
│
├── src/
│   ├── preprocessing/
│   │   └── grayscale.py
│   ├── segmentation/
│   │   └── threshold.py
│   ├── morphology/
│   │   └── operations.py
│   ├── analysis/
│   │   └── components.py
│   ├── visualization/
│   │   └── draw.py
│   └── pipeline/
│       └── pipeline.py
│
├── resources/
│   └── coins.jpg
│
└── outputs/
    └── result.png

---

⚙️ Installation

Requirements

• Python 3.x
• OpenCV
• NumPy
• PyYAML

Environment Setup & Installation

1. Create a Virtual Environment:

   python -m venv venv

2. Activate the Environment:

   • Mac / Linux:

     source venv/bin/activate

   • Windows:

     venv\Scripts\activate

3. Install Dependencies:

   pip install -r requirements.txt

---

▶️ Usage

1. Clone repository

git clone https://github.com/nguyenpd-vincent/VisionSeg.git
cd VisionSeg

2. Configure pipeline

Edit config.yaml:

preprocessing:
  grayscale: true

segmentation:
  method: otsu   # threshold | otsu | adaptive
  threshold:
    value: 120

  adaptive:
    block_size: 11
    C: 2

morphology:
  operation: close     # open | close | erode | dilate | none

  kernel:
    size: 5 # 3, 5, 7
    shape: ellipse     # square | ellipse | cross

analysis:
  min_area: 500
  max_area: 50000

3. Run

python main.py --image resources/coins.jpg

---

📊 Results

• Accurate foreground-background separation
• Noise reduction via morphology
• Detection and counting of objects

---

🔬 Experiments

Can explore:

• Different threshold values
• Kernel sizes (3x3 → 5x5)
• Morphological operations
• Otsu vs Adaptive thresholding

---

🎯 Learning Outcomes

This project demonstrates:

• Strong understanding of classical computer vision
• Ability to design modular systems
• Experience with configurable pipelines
• Practical segmentation and object detection skills