Technical University "Gh. Asachi" | Faculty of Automation and Computer Science

---

Project Overview

This MATLAB application is designed to automatically detect and blur human faces in real-time video streams or uploaded images, addressing privacy concerns in digital media. Built using MATLAB’s App Designer and Computer Vision Toolbox, the tool combines real-time image acquisition, face detection algorithms, and customizable blurring techniques into an intuitive graphical interface.

Core Objectives

• Protect privacy by anonymizing faces in images/videos.
• Demonstrate practical applications of computer vision (object detection, image processing).
• Provide an interactive platform for users to experiment with blurring parameters and shapes.

---

Technical Highlights

1. System Architecture

The application follows a pipeline workflow:

1. Image Acquisition:
   • Real-Time Mode: Captures frames from a webcam/external camera using MATLAB’s webcam object.
   • Static Mode: Processes uploaded images (JPG/PNG) or snapshots.
2. Face Detection:
   • Uses MATLAB’s vision.CascadeObjectDetector with the FrontalFaceCART model to identify faces.
   • Converts frames to grayscale to optimize Haar feature-based detection.
3. Blurring Process:
   • Applies a Gaussian filter (imgaussfilt) to detected face regions.
   • Supports two blur shapes:
     • Square: Blurs the exact bounding box of the detected face.
     • Circle: Creates a circular mask around the face center for natural blurring.
4. GUI Integration:
   • Live preview of original and processed images.
   • Interactive controls for blur level (0–40), shape selection, and live mode toggling.

2. Key Algorithms & Functions

• Real-Time Processing:
  • A timer object updates frames every 0.05 seconds (20 FPS) in live mode, calling updateLive to refresh the GUI.
  • The photoProcessing and photoProcessing2 functions handle square/circular blurring, respectively.
• Circular Blur Logic:
  • Computes the face’s center coordinates and radius from the bounding box.
  • Generates a circular mask using meshgrid and applies Gaussian blur only within the mask.

3. MATLAB Components Used

• App Designer: For building the GUI (buttons, sliders, image displays).
• Computer Vision Toolbox: For CascadeObjectDetector and face detection workflows.
• Image Processing Toolbox: For rgb2gray, imgaussfilt, and image masking.

---

Application Workflow

1. Launch the App:
   • Run app.m to initialize the GUI. The camera and timer start automatically.
2. Live Mode:
   • Toggle the LIVE switch to enable real-time blurring.
   • Adjust the blur intensity with the slider and select a blur shape.
3. Static Image Processing:
   • Capture: Click CAPTURE to take a webcam snapshot.
   • Upload: Click UPLOAD to process an existing image.
4. Exit: Click EXIT to terminate the application and release hardware resources.

---

Key Features

Feature	Description
Real-Time Blurring	Process live video at ~20 FPS with adjustable blur intensity.
Dual Blur Shapes	Choose between square (precise) or circular (natural-looking) blur effects.
Hardware Integration	Supports default and external cameras (code-configurable).
User-Friendly GUI	Intuitive controls with live previews and system time display.

---

Setup & Requirements

1. MATLAB Toolboxes:
   • Computer Vision Toolbox (mandatory for face detection).
   • Image Processing Toolbox (for blur operations).
   • Webcam Support Package (install via MATLAB Add-Ons).
2. Code Adjustments:
   • To use an external camera, modify setupLiveAcquisition in app.m:

     app.Camera = webcam('ZV-E10'); % Replace with your camera name  

3. Performance Tips:
   • Lower blur levels (<20) for smoother real-time processing.
   • Close background apps to reduce CPU load.

---

Practical Applications

• Privacy Protection: Blur faces in public datasets, social media, or surveillance footage.
• Educational Tool: Learn MATLAB’s App Designer, real-time image processing, and Haar cascades.
• Research Prototyping: Extend the code to experiment with other detectors (e.g., eye, smile).

---

Limitations & Future Work

• Current Limitations:
  • Only detects frontal faces; side profiles or occluded faces may be missed.
  • Circular blur may clip faces near image edges.
• Future Enhancements:
  • Add support for video file input/output.
  • Implement multi-shape blurring (e.g., polygons).
  • Integrate deep learning models (YOLO, SSD) for higher accuracy.

---

References

1. MathWorks, Computer Vision Toolbox Documentation [Online]. Available:
   https://www.mathworks.com/help/vision/index.html

2. MathWorks, Face Detection and Tracking Using Live Video Acquisition [Online]. Available:
   https://www.mathworks.com/help/vision/ug/face-detection-and-tracking-using-live-video-acquisition.html

3. MathWorks, vision.CascadeObjectDetector System Object [Online]. Available:
   https://www.mathworks.com/help/vision/ref/vision.cascadeobjectdetector-system-object.html

License: This project is for educational purposes. Use freely under MIT License.