Reversible Data Hiding Scheme Using Adaptive Block Truncation Coding Based on an Edge-Based Quantization Approach

Chia-Chen Lin, Ching-Chun Chang, Zhi-Ming Wang

TL;DR

This paper introduces a novel reversible data hiding scheme that embeds secret information into compressed images using an Adaptive Block Truncation Coding (ABTC-EQ) approach combined with Zero-Point Fixed Histogram Shifting (ZPF-HS). This method achieves high embedding capacity and image quality by intelligently adapting compression based on image edges and efficiently embedding data into both quantization levels and bitmaps.

What Problem Does This Paper Address?

The increasing reliance on digital communication, especially over the internet, has brought about significant security concerns. While traditional encryption scrambles data, it can sometimes leave clues. Information hiding, on the other hand, embeds secret data imperceptibly within cover media, making it a more stealthy approach. Existing information hiding schemes can be broadly categorized into spatial, frequency, and compression domains. Reversible schemes are particularly important for applications like medical imaging or military use, where the original image must be perfectly restored after data extraction.

While many reversible data hiding schemes exist, most focus on the spatial or frequency domains. However, images are often compressed before transmission to reduce file sizes. This paper addresses the need for reversible data hiding in the compression domain, specifically focusing on improving upon existing Block Truncation Coding (BTC)-based methods. Previous BTC-based methods often suffer from a "blocking effect" and may not achieve optimal embedding capacity or image quality. This work aims to overcome these limitations by proposing a more advanced compression technique (ABTC-EQ) and a refined data embedding strategy (ZPF-HS).

Methods

The proposed scheme combines two key techniques:

  • Adaptive Block Truncation Coding based on Edge-Based Quantization (ABTC-EQ): This is the core compression method.
  • Edge Detection: A Canny edge detector is first used to identify edges in the cover image.
  • Block Classification: Image blocks are then classified as either "edge-blocks" or "non-edge-blocks" based on the edge information.
  • Adaptive Quantization: * Non-edge-blocks: These are compressed using a standard bi-clustering approach (similar to traditional BTC).
  • Edge-blocks: These are compressed using a tri-clustering approach (k-means clustering) to capture more detail and complexity. This results in three quantization levels and a bitmap.
  • Benefit: This adaptive approach allows for better image quality reconstruction compared to standard BTC, especially in complex image regions.
  • Zero-Point Fixed Histogram Shifting (ZPF-HS): This is the novel data embedding technique.
  • Leveraging ABTC-EQ Output: ZPF-HS embeds secret bits into the compressed data generated by ABTC-EQ.
  • Two-Level Embedding: * Edge-blocks: Secret bits are embedded into both the quantization levels and the bitmap.
  • Non-edge-blocks: Secret bits are embedded only into the quantization levels.
  • Mechanism: ZPF-HS adapts the histogram shifting technique by defining a "zero point" (always set to "11" in this scheme) and a "peak point" (the most frequent value in the bitmap or quantization levels). Secret bits are embedded by shifting these values.
  • Benefit: This dual embedding strategy significantly increases the hiding capacity.

The overall process involves compressing the image using ABTC-EQ, classifying blocks, and then embedding secret bits into the resulting compressed codes using ZPF-HS. The extraction process reverses these steps to recover both the secret data and the original image.

Datasets

The paper uses eleven standard 512x512 grayscale test images for evaluation, including Lena, Baboon, Peppers, F-16, Fishing Boat, Girl, Gold hill, Sailboat, Tiffany, Toys, and Barbara. These are common benchmarks in image processing and steganography research.

Key Insights and Ideas

  • Edge-Aware Compression for Data Hiding: The core insight is that by intelligently adapting compression based on image content (specifically, edges), one can achieve better image quality while simultaneously creating more opportunities for data embedding.
  • Dual-Level Data Embedding: Embedding secret data into both quantization levels and bitmaps (for edge-blocks) is a powerful way to boost embedding capacity beyond what's possible with single-level embedding.
  • ZPF-HS for Efficient Embedding: The ZPF-HS technique provides a structured and reversible way to embed data into the compressed codes, ensuring that the original image can be perfectly reconstructed.
  • Solving the Blocking Effect: The ABTC-EQ approach, by considering edge information and using adaptive quantization, aims to mitigate the visual artifacts (blocking effect) often associated with traditional BTC methods.

Key Results

  • High Embedding Capacity: The proposed method demonstrates a significant embedding capacity, with an average capacity of around 74,138 bits for 4x4 blocks and 90,489 bits for 8x8 blocks.
  • Good Image Quality: The average PSNR achieved is around 36.327 dB for 4x4 blocks and 33.368 dB for 8x8 blocks, indicating good reconstruction quality after secret data extraction. This is noted to be significantly higher than conventional BTC.
  • Improved Embedding Efficiency (EF): When compared to previous BTC-based reversible data hiding schemes, the proposed method shows a higher embedding efficiency (EF), particularly when considering the size of the output code stream.
  • Effective Edge Block Handling: The ABTC-EQ method, by using tri-clustering for edge blocks, shows a much lower Mean Square Error (MSE) (55) compared to standard BTC (698) for complex blocks.

Limitations and Future Work

  • Unused Bitmap Value: The paper acknowledges that in ABTC-EQ, one specific bitmap value ("11") is not utilized. This represents a potential area for further improvement.
  • Large Code Stream Size: While the embedding capacity is high, the resulting code stream (CS) can be large, especially for edge blocks.
  • Future Work: The authors propose exploring how to utilize the unused "11" bitmap value to further enhance image quality and embed more secret information. They also suggest investigating the application of Convolutional Neural Networks (CNNs) and hyperchaos for these objectives.

Summary updated

Abstract

In this paper, we provide a novel reversible data hiding method using adaptive block truncation coding based on an edge-based quantization (ABTC-EQ) approach. We exploit the characteristic not being used in ABTC-EQ. To accomplish this, we first utilized a Canny edge detector to obtain an edge image and classify each block in a cover image into two versions, edge-block and non-edge-block. Subsequently, k-means clustering was used to obtain three quantization levels and derive the corresponding bit map while the current processing block was the case of an edge-block. Then Zero-Point Fixed Histogram Shifting (ZPF-HS) was applied to embed the secret information into compressed code. The experimental results show that our method provides a high embedding capacity for each test image and performance is better than other methods.
BibTeX
@article{lin2019reversible,
  author = {Chia-Chen Lin and Ching-Chun Chang and Zhi-Ming Wang},
  title = {{Reversible Data Hiding Scheme Using Adaptive Block Truncation Coding Based on an Edge-Based Quantization Approach}},
  journal = {Symmetry},
  year = {2019},
  doi = {10.3390/sym11060765},
}
Figure