MP3 Audio Watermarking in Images (Byte-Level and Bit-Level Methods)

This project demonstrates how to embed an MP3 audio file invisibly into a grayscale image and later recover it. Two watermarking methods are implemented:

1. Byte-level embedding: Each pixel stores one full byte of the audio data.
2. Bit-level embedding (LSB): Each audio byte is spread across 8 least-significant bits (LSBs) of consecutive pixels.

---

Features

• Encode and decode raw MP3 audio into an image without noticeable distortion.
• Support for Byte-level and Bit-level (LSB) watermarking.
• Metadata embedding for reliable recovery (start position, step, data length).
• Robust handling of grayscale images (color images are converted to grayscale automatically).
• Lossless watermark storage using BMP format (avoids JPEG compression artifacts).

---

Functions Overview

1. readMp3(dir_voise)

• Input: Path to a directory containing an .mp3 file.
• Output: A vector of raw bytes extracted from the MP3 file.

2. WM_Byte_code_coding(rMp3, dir_image, dir_img_s)

• Embeds audio bytes directly into pixels of a grayscale image.
• Metadata (start row, start column, step, length) is stored in the first row.
• Output: Watermarked image waterMarker_byte.bmp.

3. WM_Byte_code_decoding(dir_img_s)

• Extracts raw bytes from a watermarked image created by byte-level encoding.
• Output: Reconstructed audio file recovered_byte.mp3.

4. WM_Bit_code_coding(vector_voise, dir_image, dir_img_s)

• Embeds audio data at the bit level by storing each bit in the LSB of a pixel.
• Uses 32-bit metadata fields for robustness.
• Output: Watermarked image waterMarker.bmp.

5. WM_Bit_code_decoding(dir_img_s)

• Extracts embedded bits from an LSB watermarked image.
• Output: Reconstructed audio file recovered.mp3.

---

Requirements

• MATLAB R2020a or newer
• Input image in PNG format
• Audio file in MP3 format
• Output watermarked images are saved in BMP format

---

Important Notes

• Capacity Check:
  • Byte-level: Each pixel stores 1 byte. Capacity ≈ available pixels.
  • Bit-level: Each byte requires 8 pixels. Capacity ≈ available pixels ÷ 8.
• Metadata Consistency: Parameters start_r, start_c, step, and num_byte_voise must be consistent between encoding and decoding.
• File Formats: Always use BMP as the watermarked output format to avoid compression loss.
• Limitations: The watermark is not robust against image transformations (compression, resizing, filtering). This project demonstrates basic steganography, not robust watermarking.

---

Possible Extensions

• Activate and improve the capacity check for the byte-level method.
• Add key-based pseudo-random embedding for higher security.
• Implement progress reporting and error logging.
• Support for larger files via compression or segmentation.

---

License

This project is licensed under the MIT License. See LICENSE for details.

---

Attribution

Project author: Shady Nikooei - Digital Image Processing