A new statistical detector for DWT-based additive image watermarking using the Gauss-hermite expansion

Area of Science:

• Digital Image Processing
• Information Security
• Statistical Signal Processing

Background:

• Traditional discrete wavelet transform (DWT)-based image watermarking detectors use probability density functions (PDFs) with fixed parameters.
• This leads to poor statistical matching with image coefficients and suboptimal detection performance.

Purpose of the Study:

• To propose a novel detector for DWT-based additive image watermarking.
• To enhance detection accuracy and robustness against common image manipulations.

Main Methods:

• Utilizing a Gauss-Hermite expansion-based PDF for better statistical matching with image coefficients.
• Estimating parameters from higher-order moments of image coefficients.
• Deriving the decision threshold and receiver operating characteristics.

Main Results:

• The proposed detector demonstrates superior performance over Gaussian and generalized Gaussian (GG) detectors.
• Improved probabilities of detection and false alarm, along with higher efficacy.
• Enhanced robustness against compression, additive white Gaussian noise, filtering, and geometric attacks.

Conclusions:

• The Gauss-Hermite expansion-based PDF offers a more accurate statistical model for DWT coefficients.
• The proposed detector provides a significant advancement in DWT-based image watermarking security.
• This method presents a more robust solution for watermark detection in challenging conditions.