Improving the robustness of DCT-based image watermarking against JPEG compression

https://doi.org/10.1016/j.csi.2009.06.004Get rights and content

Abstract

A DCT-based image watermarking technique is proposed in this article. To improve the robustness of watermark against JPEG compression, the most recently proposed techniques embed watermark into the low-frequency components of the image. However, these components hold significant information of the image. Directly replacing the low-frequency components with watermark may introduce undesirable degradation to image quality. To preserve acceptable visual quality for watermarked images, we propose a watermarking technique that adjusts the DCT low-frequency coefficients by the concept of mathematical remainder. Simulation results demonstrate that the embedded watermarks can be almost fully extracted from the JPEG-compressed images with very high compression ratios.

Introduction

Along with the progress relating to computer hardware and software, the Internet has become the most popular channel for transmitting various forms of digital media. Since the environment of the Internet is open, the protection of digital images transmitted on the network has become an important research topic in recent years. Digital watermarking is a common technique to achieve copyright protection. For digital images, watermarking is the process of embedding significant data (watermark) into an image such that the embedded watermark can be detected or extracted later to make an assertion about the image [1], [2].

Generally, a watermarking scheme consists of three parts, the watermark, the watermark embedding stage and the watermark verification stage. The watermark embedding algorithm incorporates the watermark into the host image, whereas the verification algorithm extracts and authenticates the watermark determining the ownership of the image. Usually, the watermark is a visually recognizable logo or a set of meaningless character strings that represents the copyright of the owner or legal users. If a watermark can be extracted from an image in the verification stage, it could prove the copyright of the owner [3].

In the past few years, two kinds of techniques that are related to but different from image watermarking have been introduced in literature. One is image authentication technique and the other is image data hiding. The goal of image authentication is to verify the originality of an image by detecting malicious manipulations. However, most of the earlier methods for image authentication deal with all types of manipulations equally and unacceptably. That is these methods treat some practical manipulations such as image compression and image enhancement as attacks. Therefore, Lin and Chang proposed a DCT-based image authentication method which can prevent malicious manipulations but allow JPEG lossy compression [4]. In Lin and Chang's method, the authentication signatures for images are generated based on the invariance of the relationships between DCT coefficients at the same position in separate blocks of an image. More detailed theoretical analysis and experimental results can be found in [4]. The purpose of image data hiding is different from traditional cryptography. Cryptography concentrates on encrypting meaningful messages into meaningless data while image data hiding covers secret information with the host image as camouflage. Hiding data in images involves embedding a large amount of secret data into a cover image with minimal perceptible degradation of image quality. However, the hiding capacity for secret data and the distortion of the cover image are a tradeoff since more hidden data always result in more degradation on the cover image.

Digital watermarking has been shown as a valid solution to the problem of image copyright protection [5], [6]. There are two essential requirements for image watermarking. One is invisibility, namely the watermarked image should not be perceived the changes of the original image by human eyes. It means that the visual quality of watermarked images should not be destroyed by the embedding of watermark. The other is robustness, namely the watermark should be able to resist attacks, even if these attacks are deliberately made [7]. To improve the robustness of watermark, most of the recently proposed techniques embed watermark information into the low-frequency part of images [8], [9], [10]. Huang et al. proposed a novel DCT-based technique which embeds watermarks in the DC components of images [8]. The main idea behind this technique is that more robustness can be achieved if watermarks are embedded in DC components since DC components have much larger perceptual capacity than ac components. In addition, the authors took advantages of the feature of spatial masking (both luminance and texture masking) of the HVS to adaptively embed watermarks into images. For digitized images to be safely and efficiently transmitted on the Internet, watermarked images should be particularly robust to JPEG compression. Recently, Lin et al. proposed two kinds of DCT-based image watermarking techniques [9], [10]. These schemes perform well under general JPEG compression. However, when the watermarked images have to be compressed with higher compression ratio, the embedded watermarks may be destroyed seriously. To overcome this problem, an improved technique has been studied and proposed in this paper.

The rest of this paper is organized as follows. The proposed DCT-based watermark embedding and extracting algorithms are described in 2 Watermark embedding algorithm, 3 Watermark extracting algorithm, respectively. The relative parameters for the proposed watermarking scheme are provided in Section 4. Section 5 addresses our experimental results and discussion, especially on JPEG compression with very high compression ratios. Finally, the conclusions are given in Section 6.

Section snippets

Watermark embedding algorithm

The detailed steps of the proposed watermark embedding algorithm are given as follows.

Step 1. Torus automorphism permutation: in order to increase the security and robustness of watermark, the watermark pattern should be disarranged before embedding. Torus automorphism (TA) is an effective method to disperse a watermark equally and randomly [11]. Applying the concept of TA for scrambling the binary watermark before it is embedded into the host image offers cryptographic protection against

Watermark extracting algorithm

The steps of the proposed watermark extracting algorithm are very similar to that of watermark embedding algorithm, except for the step of watermark information extraction. The procedure for watermark extraction is quite simple and it doesn't need any assistance of the original host image. Note that two key parameters are needed in the watermark extraction procedure: the record of selected image blocks that were embedded with watermark bits and the modulus M. In addition, the parameters of

Secret keys

Secret keys are necessary in the proposed watermark embedding and extracting algorithms. Four kinds of secret keys are utilized in the proposed scheme. These secret keys should be preserved well for watermark extraction and verification.

  • 1.

    The parameters for Torus automorphism permutation: Torus automorphism permutation has been introduced in Section 2. The disarranging parameters k and m were defined in the first step of watermark embedding algorithm. These parameters are necessary when

Simulation results and discussion

The proposed scheme has been conducted on gray-level and color images, respectively. All these test images are with size 512 × 512 pixels. The watermark pattern used in the experiments is a binary image with size 32 × 32 pixels and it is given as Fig. 3. To provide objective judgment of the extracting fidelity, the similarity measurement, the Normalized Correlation (NC) value between the original watermark W and the extracted watermark W, is applied and it is defined as Eq. (8). In addition, the

Conclusions

An improved DCT-based image watermarking technique has been introduced in this paper. The watermark is embedded into a digital image, based on the concept of mathematical remainder, by modifying the low-frequency coefficients in DCT frequency domain. With the proposed scheme, the embedded watermark can successfully survive after attacked by image processing operations, especially for the JPEG compression with high compression ratio. Moreover, the watermark embedding and extracting processes are

Acknowledgements

The authors are supported by National Science Council (Taipei, Taiwan, ROC), National Dong Hwa University (Hualien, Taiwan, ROC), and National Formosa University (Hu-Wei, Yunlin, Taiwan, ROC).

References (13)

There are more references available in the full text version of this article.

Cited by (0)

View full text