Research On High-Capacity Reversible Watermarking Scheme

Digital watermarking is some labels or signatures which are embedded into digital multimedia host data by means of signal processing, exploiting the characteristics of Human Visual System (HVS) or Human Auditory System (HAS). The labels or signatures usually cause unnoticeable perceptual distortion to the original data and are retrieved only by an appropriate extraction algorithm. It plays a very important role in the applications to copyright protection of digital media, secret communication on network and data hiding.For some critical applications such as the fields of the law enforcement, medical and military image systems, it is crucial to restore the original image without any distortions after the watermark is extracted. The watermarking techniques satisfying these requirements are referred to as reversible watermarking. A considerable amount of research on it has been done over the past few years.In this dissertation, we firstly summarize a systematical introduction of data hiding; then mainly focus our research on the high-capacity reversible watermarking algorithm. The valuable research fruits can be summarized as follows:(1) A reversible data hiding scheme based on mixed integer transform is presented in chapter 3. In a grayscale image, every 3×3 adjacent pixels are grouped into an ennead. A nine-pixel block integer transform is applied to each ennead to get its transformed one. After that, the companding technique is employed to compress its eight difference values. Then, data embedding is carried out by expanding the smaller differences. The introduction of the companding technique is largely increase the number of enneads available for embedding. For the non-expanded ennead, a three-pixel block integer transform is introduced to embed two bits into the 3×1 triplet. Because the location map recording these expanded positions can be efficiently compressed due to the mixed integer transform, a high hiding capacity is achieved. Our method can achieve embedding rate of up to 8/9 bpp at once embedding.(2) Reversible watermarking based on multi-scale and difference compression is proposed in chapter 4. The embedding process based on invariability of the sum of pixel pairs is carried out by using difference compression of pixel pairs. For every pixel is utilized twice, a very high capacity is achieved. The capacity consumed by the compressed location map is largely decreased because of the embedding order, which is performed from the larger scale to the smaller one. Our algorithm can achieve embedding rate of close to 1 bpp without multiple embedding.