Reversible Authentication Watermarking Based on Improved 2D Histogram and Adaptive Difference Expansion

Abstract

To address the limitations of low authentication accuracy and ineffective protection for complex-texture images/regions in existing reversible schemes, an improved algorithm based on two-Dimensional (2D) histogram and difference expansion is proposed. The core innovation involves classifying image pixels into texture and smooth categories using local complexity analysis. Distinct embedding strategies are then applied: texture pixels utilize adaptive prediction difference expansion for authentication watermark embedding, while smooth pixels utilizes channel shifting within the improved 2D histogram for authentication watermark embedding. Furthermore, a hierarchical embedding strategy enhances both authentication effectiveness and visual quality per sub-block. Experimental results demonstrate that the method achieves an average tamper detection rate of 92.78% (using 16×16 blocks) under complex attacks while maintaining PSNR above 46 dB. Compared to state-of-the-art methods, it significantly improves tamper localization for complex-texture content without compromising reversibility or visual fidelity.