Capacity estimates for data hiding in compressed images

Area of Science:

• Digital Image Processing
• Information Theory
• Data Security

Background:

• Data-hiding in images is crucial for digital security.
• Transform domains are commonly used for embedding, but the optimal transform is unclear.
• Existing methods often use Discrete Cosine Transform (DCT) or wavelets.

Purpose of the Study:

• To estimate the maximum number of bits that can be hidden in still images (data-hiding channel capacity).
• To investigate the impact of embedding data in transform domains versus the spatial domain.
• To compare the data-hiding capacities of various transform domains.

Main Methods:

• An information-theoretic approach was used to estimate channel capacity.
• Data embedding was performed in different transform domains, including DCT, Discrete Fourier Transform (DFT), Hadamard, and subband transforms.
• Comparisons focused on achievable data-hiding capacities for each transform.

Main Results:

• Embedding data in a suitable transform domain significantly increases channel capacity compared to the spatial domain.
• The magnitude DFT decomposition demonstrated the highest data-hiding capacity among the tested transforms.
• The choice of transform domain critically affects the achievable data-hiding capacity.

Conclusions:

• The magnitude DFT offers superior performance for data-hiding channel capacity estimation.
• Optimizing the transform domain is essential for maximizing data embedding in images.
• This research provides a framework for selecting optimal transforms for secure data-hiding.