Source-linked AI summary
Reversible data hiding based on reducing invalid shifting of pixels in histogram shifting
Yujie Jia, Zhaoxia Yin, Xinpeng Zhang, Yonglong Luo
TL;DR
Existing reversible data hiding methods often overlook image texture, even though invalid shifting pixels affect embedding distortion. The proposed scheme partitions images in a checkerboard pattern, calculates fluctuation values, and prioritizes lower-fluctuation regions; experiments report reduced distortion and increased capacity, with performance differences between smooth and texture images left for future study.
Problem
Existing RDH methods do not fully account for image texture and the role of invalid shifting pixels in embedding distortion.
Method
The scheme checkerboard-partitions the image, calculates and sorts pixel fluctuation values, and combines them with prediction errors to embed data preferentially in smoother regions.
Results
The proposed scheme reports fewer invalid shifting pixels, reduced distortion, improved capacity, and better performance than several existing schemes across tested images and payloads.
Takeaways & Limitations
Texture-aware prioritization links smoother embedding regions with fewer invalid shifts and supports improved stego-image quality and embedding capacity within the tested setting.
Takeaways & Limitations
The authors identify further study of performance differences between smooth and texture images and development of separate reduction methods as future work.
Abstract
from arXiv · showhide
In recent years, reversible data hiding (RDH), a new research hotspot in the field of information security, has been paid more and more attention by researchers. Most of the existing RDH schemes do not fully take it into account that natural image's texture has influence on embedding distortion. The image distortion caused by embedding data in the image's smooth region is much smaller than that in the unsmooth region, essentially, it is because embedding additional data in the smooth region corresponds to fewer invalid shifting pixels (ISPs) in histogram shifting. Thus, we propose a RDH scheme based on the images texture to reduce invalid shifting of pixels in histogram shifting. Specifically, first, a cover image is divided into two sub-images by the checkerboard pattern, and then each sub-image's fluctuation values are calculated. Finally, additional data can be embedded into the region of sub-images with smaller fluctuation value preferentially. The experimental results demonstrate that the proposed method has higher capacity and better stego-image quality than some existing RDH schemes.
1. Introduction
RDH enables both data extraction and perfect cover-image recovery, but existing approaches often overlook how image texture and invalid shifting pixels affect distortion. The paper motivates texture-aware embedding through histogram-shifting and prediction-error methods.
- Reversible data hiding extracts embedded data while allowing the original cover image to be reconstructed without distortion.
- Plaintext-domain RDH embeds additional data directly into unencrypted cover images, alongside encrypted-domain approaches for encrypted images.
- Existing RDH methods use lossless compression, difference expansion, histogram shifting, and prediction-error expansion, with histogram shifting widely used.
- Histogram shifting embeds data through peak-point pixels and shifts intervening pixels, but its capacity is relatively low.
- Many methods scan pixels without considering texture, although more invalid shifting pixels produce greater image distortion.Invalid shifting pixels are shifted but not selected for embedding; for fixed payloads, valid shifting is fixed, so distortion depends substantially on invalid shifting.
- The proposed scheme calculates fluctuation values after checkerboard partitioning and preferentially embeds data in lower-fluctuation regions.The method combines prediction error with fluctuation values to prioritize smoother image regions.
2. Proposed method
The proposed method embeds data preferentially in smoother regions identified through checkerboard sub-images and fluctuation values, while using prediction errors and reversible histogram shifting for embedding and recovery.
- 2.1. Calculation of fluctuation value: The cover image is divided into checkerboard sub-images A and B, whose fluctuation values guide region selection.The two sets contain gray and white checkerboard pixels, and their fluctuation calculations use the same procedure.
- 2.1. Calculation of fluctuation value: Each pixel’s fluctuation value is computed from local complexity and the complexity of adjacent pixels.Local complexity uses neighboring pixels in vertical, horizontal, and diagonal directions; boundary cases use one, two, or four adjacent pixels.
- 2.2. Calculation of prediction error: Prediction errors are calculated separately for the two layers using four nearest neighbors and weighted prediction.The weights correspond to the upper, lower, left, and right neighbors and sum to 1.
- 2.3. Embedding procedure: The method prevents underflow and overflow with a lossless location map and reverses embedding by extracting and recovering B, then A.Boundary values 0 and 255 are temporarily changed to 1 and 254, while the location map is compressed as part of the payload.
- 2.3. Embedding procedure: Pixels are processed in ascending fluctuation order, and half of the payload is allocated to each layer for prediction-error embedding.The embedding rules modify errors at peak points and shift errors between peak and zero points; the second layer is then processed similarly.
- 2.4. Extraction and recovery procedure: During extraction, B’s fluctuation values remain unchanged because they are computed from A, enabling the original embedding order to be reconstructed.Marked prediction errors are sorted according to ascending fluctuation values before data extraction and recovery.
3. Experimental results and analysis
Experiments on six 512 × 512 test images compare the proposed method with three schemes using image quality and embedding capacity. The proposed method reduces ISPs and distortion while improving PSNR and capacity across smooth and textured images.
- Experimental setup: Experiments use six standard 512 × 512 images and compare the proposed method with Chen et al., Lu et al., and Jung K H.Evaluation targets visual quality and embedding capacity.
- Experimental setup: The method preferentially embeds data at low-fluctuation positions, reducing ISPs and improving performance through texture-aware prediction-error histogram embedding.This design targets smooth pixels, where fewer invalid shifts are produced.
- Comparison of the number of ISPs: 3,220 ISPs occur for Airplane at 10000 bits, 22600, 14369, and 15506 fewer than the three comparison methods, respectively.For Baboon, the proposed method has fewer ISPs than half of the three comparison methods.
- Comparison of PSNR: At 0.03 bpp on Airplane, PSNR is 64.45 dB for the proposed method versus 58.97, 59.64, and 59.49 dB for the comparison schemes.The proposed method improves PSNR by 5.48, 4.81, and 4.96 dB, respectively.
- Comparison of PSNR: Average PSNR gains over 0.01∼0.05 bpp are consistent across test images, reaching 6.76 dB over Jung K H on Man.The method reduces distortion on both smooth and unsmooth images.
- Comparison of embedding capacity: On Baboon, embedding capacity is 25038 bits for the proposed method versus 18371, 24997, and 16271 bits for the three comparison schemes.The corresponding increases are 36.29%, 0.16%, and 53.88%; on Lena, increases are 38.83%, 0.33%, and 96.75%.
4. Conclusions
The proposed RDH scheme combines image texture with checkerboard partitioning to reduce invalid shifting pixels and improve embedding performance. Its reported advantages include higher PSNR and increased embedding capacity, while future work remains for smooth and non-smooth images.
- The scheme partitions images using a checkerboard pattern and embeds data preferentially in smoother pixels to reduce invalid shifting.It calculates and sorts fluctuation values to identify smoother regions; checkerboard partitioning also reduces auxiliary information.
- Compared with three state-of-the-art schemes, the proposed method significantly improves PSNR and visual quality.
- Selecting double peaks substantially increases the proposed scheme’s embedding capacity.
- Future work will investigate separate methods for reducing invalid shifting in smooth and non-smooth images.