International Journal of Innovative
Computing, Information and Control ICIC International c°2008 ISSN 1349-4198
Volume 4, Number 2, February 2008 pp. 351—358
A NOVEL REVERSIBLE DATA HIDING SCHEME
Shaowei Weng, Yao Zhao
Institute of Information Science Beijing Jiaotong University Beijing 100044, P. R. China
[email protected]; [email protected]
Jeng-Shyang Pan
Department of Electronic Engineering Kaohsiung University of Applied Sciences 415 Chien Kung Road, Kaohsiung City 80778, Taiwan
Received November 2006; revised April 2007
Abstract. A novel reversible data hiding scheme based on an improved difference ex-pansion (DE) method, which introduces the companding technique into traditional DE, is proposed in this paper. The method focuses on largely increasing the number of the differences to be expanded. The location map recording these expanded positions can be efficiently compressed due to significant difference between possibilities for expanded and non-expanded differences. Thereby, the hiding capacity is largely increased. We compress large differences into smaller values, and then expand these values. Hence, the higher PSNR value can be achieved than that caused by directly expanding large differences. As revealed in the experimental results, the proposed scheme outperforms traditional DE, in terms of embedding capacity and visual quality, as expected.
Keywords: Reversible watermarking, DE, Companding technique
1. Introduction. Digital watermarking refers to the process of embedding some labels or signatures into digital media without introducing perceptible artifacts. Watermark-ing schemes can be classified as either robust or fragile. Robust watermark is generally designed to resist unmalicious or malicious attacks such as scaling, cropping, lossy com-pression, and so forth. For example, Lu et al. proposed two counterfeiting attacks on two robust watermarking schemes [1]. In contrast, fragile watermark is designed to detect any tiny alternation to the original digital content. As one kind of fragile watermark, re-versible watermarking has drawn extensive attention from researchers especially in some fields such as the law enforcement, medical and military image systems, as it can restore the original image without any distortions after the watermark is extracted.
The concept of reversible watermarking firstly appeared in the patent owned by East-man Kodak [2]. Honsinger et al. utilized a robust spatial additive watermark combined with modulo addition to achieve reversible data embedding [2]. Fridrich et al. losslessly compressed high level bit-planes to make space for the embedded data in the spatial do-main [3]. Macq employed the patchwork algorithm, proposed by Bender et al. [4], and modulo addition to reversibly embed the watermark bits into the host image [5]. The above algorithms [2, 3, 5] aimed at the authentication, so their amounts of hiding data were limited. The following algorithms [6, 7, 8, 9, 10] can achieve higher hiding capacity. Goljan et al. proposed a two cycle flipping permutation to assign a watermark bit in each pixel group [6]. Celik et al. presented a high capacity, reversible data-embedding algo-rithm with low distortion by compressing quantization residues [7]. Xuan et al. embedded
