Reversible data hiding is essential for data security and integrity in industrial environments. However, existing methods suffer from unstable prediction errors and limited embedding capacity. This article proposes an adaptive reversible data hiding framework for industrial images using enhanced content pixel prediction error. Initially, a refined content pixel selection method is proposed to accurately identify embeddable pixels in industrial images. Then, a distance-based pixel prediction scheme is developed that utilizes high-relevance content pixels for improved prediction accuracy. In addition, a content-adaptive embedding strategy is introduced that prioritizes smooth blocks while employing prediction error pair mapping to simultaneously maximize embedding capacity and minimize distortion. Experimental results show that the proposed method outperforms state-of-the-art approaches in embedding capacity, achieving an average of 45 625 bits per image with the capability to embed in complex textured regions. Extensive evaluation on standard benchmarks and two industrial image databases validates its effectiveness for secure industrial data transmission.

Abstract Adaptive embedding plays an important role in improving the embedding performance of reversible data hiding and it is usually realized by modifying the prediction‐errors discriminately. By extending the skewed histogram shifting technique which uses a pair of extreme predictions to determine whether the target pixel should be predicted or not, this paper realizes another form of adaptive embedding, that is, adaptive prediction‐error generation. Specifically, it is proposed to adaptively determine the pair of extreme predictions according to image content. Each pair of extreme predictions is first evaluated by the introduced distortion per embedding one bit. Then, an efficient mechanism to determine the best pair of extreme predictions for pixels with a given local complexity is designed. With such a mechanism solving the computational problem, it is also proposed to extend the context to obtain more pairs of extreme predictions such that more precise multi‐predictor can be realized. With obtained errors, the best expansion bins are determined to achieve a more comprehensive self‐adaption. Experimental results demonstrate that the proposed scheme achieves better capacity‐distortion performance and outperforms a series of state‐of‐the‐art schemes.

Recent reversible data hiding (RDH) work tends to realize adaptive embedding by discriminately modifying pixels according to image content. However, further optimization and computational complexity remain great challenges. By presenting a better incorporation of pixel value ordering (PVO) prediction and pairwise prediction-error expansion (PEE) technologies, this paper proposes a new RDH scheme. The largest/smallest three pixels of each block are utilized to generate error-pairs. To achieve optimization of the distribution of error pairs, two-layer embedding is introduced such that full-enclosed pixels of each block can be used to determine how to optimally define the spatial location of pixels within block. Then, to modify error pairs with less distortion introduced, the shifted pairing error is involved in the separable utilization of the other one; i.e., it serves as the context for recalculating the other one. Since the recalculation is equivalent to expansion bins selection, various extensions of original pairwise PEE are designed, parameterized, and combined into the so-called multiple pairwise PEE, with which the 2D histogram can be divided into a set of sub-ones for more accurate modification. The experimental results verify the superiority of the proposed scheme over several PVO-based schemes. On the Kodak image database, the average PSNR gains over original PVO-based pairwise PEE are 0.83 and 0.99 dB for capacities of 10,000 and 20,000 bits, respectively.

Reversible data hiding in encrypted image (RDHEI) is a technique that can be adopted by cloud sever to embed additional data into the encrypted image with no permanent distortion. For RDHEI, it remains a challenging task to improve the embedding capacity under the premise of real reversibility. In this paper, a novel RDHEI method based on adaptive most significant bit (MSB) prediction is proposed. The cover image is first encrypted in block-wise manner such that the correlation of pixels within block is preserved. Next, all blocks are permuted to fulfill the final encryption. During data embedding, the upper-left pixel within block is used to predict others such that the embedding room is vacated. Then, available blocks are selected and all blocks are rearranged so as to ensure reversibility. By fully exploiting the correlation of pixels within block via adaptive MSB prediction, the proposed method successes to achieve desirable improvement in capacity. Experimental results show that the proposed method significantly outperforms previous methods. Moreover, real reversibility and real separability are also guaranteed. In a word, the proposed method is a practical method that can be adopted by cloud storage.

Pixel value ordering prediction has been verified as an effective mechanism to exploit image redundancy for reversible data hiding (RDH) and numerous extensions have been devised. However, their performance is still unsatisfactory since the error modification is generally fixed and independent of image content. In this paper, a new RDH scheme is proposed by incorporating pixel distance to realize adaptive multiple histograms modification (AMHM). During exploiting the correlation between the largest/smallest pixel and any other one in the scope of pixel block, we propose to process every two correlated pixels successively following the ascending order of their distance. Specifically, the generated errors with a given distance are collected and verified. If they are all shiftable errors, the follow-up errors would be collected into the next sub-histogram. In this way, a histogram sequence is adaptively generated such that different modification mechanisms can be taken for different sub-histograms to achieve adaptive embedding. Finally, AMHM for conventional prediction-error expansion (PEE) and AMHM for 2D PEE have been both realized in this paper. Experimental results show that AMHM is of great significance to better exploit pixel correlation and the proposed scheme outperforms a series of the latest schemes.

Reversible data hiding generally exploits the redundancy of the cover medium and prediction-error expansion (PEE) has become the most effective mechanism. However, although the pairwise PEE technique has been proposed to jointly modify the prediction-errors to achieve less degradation, there is still room for improvement. In this paper, a dual pairwise PEE strategy is proposed to fully exploit the potential of pairwise PEE. The key observation behind dual pairwise PEE lies in that most capacity is provided by individually expanding only one pairing error. For such separable error-pairs, we propose to recalculate and collect the rest pairing error to form an error sequence after shifting any one pairing error. Next, by considering every two neighboring errors of the sequence together, a new set of error-pairs for double pairwise PEE can be obtained. Compared with original pairwise PEE, dual pairwise PEE significantly better exploits the correlation of errors such that it leads to better capacity-distortion performance. Experimental results also demonstrate that the proposed scheme outperforms several state-of-the-art schemes.

In reversible watermarking for image authentication, less degradation of the marked image is always desirable. For minimum distortion, the pairwise prediction-error expansion (PEE) technique was recently proposed to modify errors jointly. Although its superiority over conventional PEE has been verified, its potential has not been fully exploited yet. In this paper, we focus on optimal modification and propose an enhanced pairwise PEE. First, it is observed in PVO-based pairwise PEE that the histogram peak varies with relative location when predicting the largest/smallest two pixels. Then, a more effective 2D mapping is proposed by content-dependently selecting the expansion bin after introducing spatial location into prediction. Next, the 2D mapping is further extended considering prediction in non-smooth region tends to produce errors with large magnitude. Finally, we also propose to flexibly define the spatial location to achieve content-dependent prediction and further enhancement. Experimental results demonstrate that the proposed scheme achieves better capacity-distortion trade-off and outperforms several state-of-the-art schemes.

The cover image is generally encrypted by a stream cipher in existing reversible data hiding in encrypted image (RDHEI) methods. As pixel correlation is seriously damaged, more than one pixel should be employed to carry one bit such that the quite limited capacity is achieved. To overcome this issue, a new RDHEI method with high capacity, that preserves pixel correlation and exploits it to vacate embedding room, is proposed in this paper. First, we propose a block-level encryption scheme which combines block-level stream cipher and block-level permutation, and all blocks are classified into usable blocks (UBs) and unusable blocks (NUBs) by preserving the correlation of pixels in blocks. Then, UB is reconstructed to vacate room for data embedding, because the pixels in blocks share the same most significant bits (MSBs). To ensure reversibility, the number of NUBs between current UB and the previous one is also embedded along with additional data, and the blocks are rearranged in a reversible way such that UBs are always in front of NUBs. Experimental results show that not only the embedding capacity is significantly improved but also the hidden data can be losslessly extracted, and the cover image can be perfectly recovered.

As the core of prediction-error expansion technique, prediction method has a fundamental impact on performance of reversible data hiding. Pixel value ordering (PVO) prediction has been extensively investigated for its high accuracy. However, the correlation of pixels within block has not been fully exploited yet. In this paper, a novel prediction method is proposed by developing PVO prediction in the aspects of spatial correlation and correlated pixel pair. The key to PVO embedding is invariant pixel value order such that the predicted pixel can be identified by value. Instead of predicting and enlarging the largest pixel, we propose to predict and reduce the second largest one and even all others. As the largest pixel which serves as predicted value is maintained after embedding, the numerous predicted pixels can be identified and thus reversibility is guaranteed. IPVO prediction which location-dependently determines the predicted pixel is also developed. For further optimization, multi-pass IPVO embedding is extended from single-layered to double-layered such that full-enclosing pixels can be used to estimate pixel distribution and determine the optimal mode of defining spatial location. Finally, an enhanced pairwise PEE is incorporated with multi-pass IPVO for performance enhancement. Experimental results show that the proposed scheme not only outperforms PVO embedding and its miscellaneous extensions, but also achieves significant superiority in fidelity over a series of state-of-the-art schemes.