🎯 High-Precision Face Detection using MATLAB GUI

📌 Overview

This project presents a face detection system developed in MATLAB as part of the Image Processing (ECE228) course at Zagazig University. The system is capable of accurately detecting human faces in digital images by leveraging a custom-designed algorithm that avoids the use of high-level built-in face detection functions.

Instead, we focused on implementing fundamental image processing techniques such as skin tone segmentation, symmetry analysis, and geometric filtering to detect facial regions. The project also includes a full Graphical User Interface (GUI) to allow users to interactively load images, start detection, and view results.

---

🎯 Objectives

• Develop a face detection system from scratch using basic image processing techniques.
• Avoid the use of built-in high-level detectors (like vision.CascadeObjectDetector).
• Build a GUI to support non-technical users in running detection easily.
• Ensure accurate detection under different lighting conditions and backgrounds.

---

🛠️ How It Works

The system pipeline consists of the following main stages:

1. Image Preprocessing

   • Apply Gaussian filter to reduce noise.
   • Convert the RGB image to YCbCr color space to better separate skin tones.

2. Skin Detection

   • Threshold the Cr and Cb channels to isolate potential skin regions.
   • Generate a binary skin mask.

3. Morphological Processing

   • Clean up the binary mask using morphological operations:
     • imfill to fill holes
     • bwareaopen to remove small objects

4. Region Analysis

   • Extract connected components using regionprops.
   • Measure area, eccentricity, and bounding boxes.

5. Face Verification

   • Verify potential face regions using:
     • Elliptical shape matching
     • Horizontal symmetry (left/right pixel similarity)
     • Eccentricity constraint to rule out elongated regions

---

🖥️ GUI Features

• Load Image: Import an image from the local machine.
• Start Detection: Run the detection pipeline and draw bounding boxes around detected faces.
• Reset: Clear the interface for a new image.
• Exit: Close the application.

---

✅ Results & Performance

• Accuracy: High for well-lit, frontal face images.
• False Positives: Very low due to strict shape and symmetry checks.
• Speed: Acceptable for real-time use within the GUI.
• Robustness: Performs well with variations in skin tone and lighting.

---

⚙️ Requirements

• MATLAB R2021a or newer
• Image Processing Toolbox

---

📃 License

This project was developed for academic purposes and is open for educational use.
Not intended for commercial applications.

---

📂 Repository Structure