Image watermarking in Python

Description

Image watermarking is one of the most used methods for hiding an image in another image.

A comparative study between DCT, DWT and DFT in terms of watermarking :

Discrete Fourier Transform

Advantages :

•  Rotational resistant
• Translation Invariant
• High robustness
• Performs best under geometrical attacks

Disadvantages :

1. Due to approximation through the process the accumulation of error increases.
2. Loss of quality while extraction process
3. Time taken is also much higher

 

Discrete Cosine Transform

Advantages :

• Since the image is decomposed into multiple levels here the middle frequency level can be used to embed the required watermark which makes it more resistant towards attacks that
  affect the low or high frequency
• Compared to spatial transforms this is more robust in nature

Disadvantages:

•  Loss in quality of the image when compressed

 

 

Discrete Wavelet Transform

Advantages :

• Spatio-frequency localization is better.
• Increased Versatility.
• High imperceptibility.
• Can carry more watermark with decent concealing effect

Disadvantages :

• Adding a watermark might degrade the image.
• The implementation cost is very high.

 

Singular Value Decomposition

Advantages :

• Produces the least mean square truncation error.
• Deals in matrix level.
• Highly Stable.
• Uses variable orthogonal basis hence the error is very low

Disadvantages :

• Sensitive to noise in general
• Low Distortion is experienced here

 

Contents :

In this python project, we implement four methods of image watermarking.

DWT (Discrete wavelet transform)

DFT ( Discrete Fourier transform )

DCT (Discrete cosine transform)

SVD ( Singular value decomposition )

and two-hybrid methods :

DWT_SVD (Discrete wavelet transform and Singular value decomposition)

DWT_DCT_SVD (Discrete wavelet transform and Discrete cosine transform and Singular value decomposition)

All the codes were tested in python 3.