Dynamic Hybrid Reversible Data Hiding Based on Pixel-value-ordering

Abstract

Reversible data hiding (RDH) using pixel-value-ordering (PVO) is a well-established technique for embedding data in a cover image by modifying the maximum and minimum in each block. This paper proposes a dynamic hybrid RDH method based on PVO. Specifically, a 3×3 block according to its complexity and two thresholds T1 and T2 is classified as three levels: extremely smooth, smooth, and rough. Different processing algorithms are used for different levels. For rough blocks, they are ignored to avoid reducing the peak signal-to-noise ratio. For smooth blocks, the proposed method employs a block subdivision algorithm that can embed up to 6 bits of data. For extremely smooth blocks, no subdivision is done and a median pixel prediction algorithm is used to predict the remaining eight pixels, which can embed up to 8 bits of data. Moreover, this paper presents a new method that computes complexity by dynamically selecting relevant pixels to enhance block classification accuracy. Extensive experiments demonstrate that the proposed method outperforms existing PVO-based methods, offering larger embedding capacity while maintaining low distortion.