[论文解读] Robust LSB watermarking optimized for local structural similarity

Amin Banitalebi Dehkordi, Said Nader Esfahani+1 more|arXiv (Cornell University)|May 17, 2011

Advanced Steganography and Watermarking Techniques参考文献 16被引用 21

一句话总结

本文提出了一种针对结构相似性（SSIM）优化的自适应LSB水印方案，提升了鲁棒性与容量，同时保持了不可感知性。通过根据SSIM阈值动态逐块嵌入水印，该方法在性能上优于传统的基于MSE的方案，在各种攻击下仍能保持高提取精度。

ABSTRACT

Growth of the Internet and networked multimedia systems has emphasized the need for copyright protection of the media. Media can be images, audio clips, videos and etc. Digital watermarking is today extensively used for many applications such as authentication of ownership or identification of illegal copies. Digital watermark is an invisible or maybe visible structure added to the original media (known as asset). Images are considered as communication channels when they are subject to a watermark embedding procedure so in the case of embedding a digital watermark in an image, the capacity of the channel should be considered. There is a trade-off between imperceptibility, robustness and capacity for embedding a watermark in an asset. In the case of image watermarks, it is reasonable that the watermarking algorithm should depend on the content and structure of the image. Conventionally, mean squared error (MSE) has been used as a common distortion measure to assess the quality of the images. Newly developed quality metrics proposed some distortion measures that are based on human visual system (HVS). These metrics show that MSE is not based on HVS and has a lack of accuracy when dealing with perceptually important signals such as images and videos. SSIM or structural similarity is a state of the art HVS based image quality criterion that has recently been of much interest. In this paper we propose a robust least significant bit (LSB) watermarking scheme which is optimized for structural similarity. The watermark is embedded into a host image through an adaptive algorithm. Embedding will continue in each block of the host image until a certain level of threshold in the structural similarity is met. Various attacks examined on the embedding approach and simulation results revealed the fact that the watermarked sequence can be extracted with an acceptable accuracy after all attacks. In comparison to the original algorithm, the proposed method increases the capacity while the imperceptibility and robustness remain fix.

研究动机与目标

• 解决传统基于MSE的失真度量在评估水印图像质量方面的局限性。
• 通过将SSIM作为质量度量，使水印性能与人类视觉系统（HVS）感知对齐，从而提升性能。
• 在LSB基水印中优化不可感知性、鲁棒性与嵌入容量之间的权衡。
• 开发一种自适应嵌入算法，根据局部图像结构动态调整，以维持SSIM阈值。

提出的方法

• 采用结构相似性（SSIM）作为主要失真度量，而非均方误差（MSE），更准确地反映人类视觉感知。
• 将宿主图像划分为块，并计算局部SSIM值，以在嵌入前评估感知影响。
• 应用自适应LSB水印策略，逐块嵌入数据，直至达到预设的SSIM阈值。
• 通过在过程中监控SSIM水平，确保水印嵌入不会降低感知质量。
• 采用基于阈值的停止条件，以控制嵌入强度并保护图像结构。
• 通过在各种攻击下的仿真评估鲁棒性，确认了水印的可靠提取。

实验结果

研究问题

• RQ1 SSIM能否作为比MSE更准确且更符合感知相关性的度量，用于优化LSB水印？
• RQ2 基于局部SSIM的自适应嵌入如何改善不可感知性与容量之间的平衡？
• RQ3 所提出方法在常见图像处理攻击下能多大程度上保持鲁棒性？
• RQ4 通过SSIM优化是否可在不损害视觉质量的前提下实现更高的嵌入容量？

主要发现

• 所提方法相比传统LSB方案提高了水印嵌入容量。
• 不可感知性得以保持，因为算法确保嵌入过程中SSIM始终高于预设阈值。
• 在各种攻击下仍保持了鲁棒性，仿真中报告了准确的水印提取。
• 该方法通过与人类视觉系统特性对齐，优于传统的基于MSE的方案。