🖼️ Foreground Extraction & Mask-Based Blending

📌 Project Overview

This project demonstrates the fundamentals of Digital Image Processing (23DSDE12) through mask-based foreground extraction and image compositing. Using a transparent image of a character (Tom/Jerry), the script dynamically generates a binary mask and blends the foreground onto a new cartoon background scene.

The repository provides two parallel implementations:

1. MATLAB (experiment1.m): The primary academic implementation.
2. Python/OpenCV (experiment1.py): A production-grade implementation mirroring the exact MATLAB logic.

🚀 Features

• Automatic Mask Generation: Automatically detects and extracts the alpha channel of transparent images (.png) to create flawless binary masks without manual thresholding.
• Robust Path Resolution: Compatible with both local VS Code execution and MATLAB Online's virtual file system.
• Cross-Language Validation: Verifies results by implementing the exact mathematical operations in both MATLAB and Python.
• Automated Directory Management: Automatically generates the output/ directory and saves processed step-by-step images.

📂 Directory Structure

📦 Foreground-Extraction
 ┣ 📂 input
 ┃ ┣ 📜 tom.png            # Foreground object (with alpha channel)
 ┃ ┣ 📜 jerry.png          # Alternate foreground object
 ┃ ┗ 📜 background.jpg     # Background scene
 ┣ 📂 output
 ┃ ┣ 📜 composite.png      # Final blended image
 ┃ ┣ 📜 mask.png           # Extracted binary mask
 ┃ ┣ 📜 extracted_fg.png   # Foreground cut-out
 ┃ ┣ 📜 masked_bg.png      # Background with hole punched
 ┃ ┗ 📜 Experiment1_Results.png # 2x3 Subplot showing all steps
 ┣ 📜 experiment1.m        # MATLAB implementation
 ┣ 📜 experiment1.py       # Python/OpenCV implementation
 ┗ 📜 README.md

🛠️ Usage

MATLAB Installation

1. Open MATLAB Desktop or MATLAB Online.
2. Navigate to this repository's directory.
3. Open experiment1.m and click Run.
4. The output subplot will render, and files will be saved in output/.

Python Installation

# Install dependencies
pip install opencv-python numpy matplotlib

# Run the pipeline
python experiment1.py

🧠 Methodology

1. Read & Preprocess: Loads images and extracts the alpha transparency channel to use as a master mask.
2. Resize: Rescales the background matrix and the mask matrix to perfectly match the foreground object dimensions using bicubic interpolation.
3. Binarization & Precision: Normalizes all image matrices to double-precision [0, 1] to ensure pixel-perfect array multiplication.
4. Compositing:
   • Extracted Foreground = fg .* mask
   • Masked Background = bg .* (1 - mask)
   • Final Output = Extracted Foreground + Masked Background

👤 Author

Sanyog Kumar Singh
USN: 23BTRDC034
Subject: Digital Image Processing (23DSDE12)