The Study of Image Steganography and Digital Watermarking

• Main Authors: Chi-Shiang Chan, 詹啟祥
• Other Authors: Chin-Chen Chang
• Format: Others
• Language: en_US
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/59385727269146539154

博士 === 國立中正大學 === 資訊工程所 === 94 === There are two topics in this dissertation. One is image steganography and the other is digital watermarking. Although image steganography and digital watermarking are two different topics, there exists one common point. Both image steganography and digital watermarking embed data into images except that digital watermarking has an additional property, the robustness of the embedded data. Therefore, we can treat digital watermarking as a special case of image steganography. Owing to this common point, two topics are studied in this dissertation. Regarding, image steganography, we consider the problems of gray-level image hiding and color image hiding. For gray-level images, we propose three different methods; a color image hiding technique is also proposed for color images. Among all related works for gray-level images, least-significant-bit (LSB) substitution is the most straightforward way to embed a secret image into a host image. To avoid the image degradation of the LSB substitution technique, Wang et al. proposed a method using the substitution table to perform image hiding. In 2001, Wang et al. brought up the concept of one-to-one mapping where each secret data value is transformed into another value by applying the table. Nevertheless, Wang et al. have not developed any method to obtain an optimal one-to-one substitution table. In 2003, Chang et al. proposed a way to find an optimal one-to-one substitution table by using the dynamic programming strategy. In this dissertation, we have proposed three gray-level image hiding techniques by using an optimal multi-bit-reference substitution table, an optimal one-to-two substitution table and an optimal modulus function based substitution table. In the first method, we use un-embedded bits of host pixels to partition the host pixels into different planes. We can derive an optimal substitution table for each plane. After combining all optimal substitution tables, we can obtain an optimal multi-bit-reference substitution table. The secret data being transformed via the optimal multi-bit-reference substitution table will degrade the host image possibly least. In the second method, we introduce a new method that uses an optimal one-to-two substitution table. We first built up a square-error matrix with 3-dimension. After applying a dynamic programming strategy on the square-error matrix, we can obtain an optimal one-to-two substitution table. Combing the modulus function, proposed by Thien and Lin, and the optimal substitution table, we develop a novel method to improve the image quality of the stego-image. Now, consider the problem of color image embedding. The secret image to be embedded is first color-quantized using a palette previously sorted by the principal component analysis technique to generate the indexed color image. Then, the concept of pixel difference is employed to process both the host image and the indexed color image. After that, the processed indexed color image is embedded into the processed color host image. Finally, we describe the second topic, digital watermarking, in this dissertation. Digital watermarking is used for the copyright protection of digital images. To embed the watermark, the proposed method partitions an original image into blocks and uses a PCA function to project these blocks to a linear subspace. There is a watermark table, which is computed from projection points, kept in our proposed method. When extracting a watermark, the proposed method projects the blocks of the modified image by using the same PCA function. Both the newly projected points and the watermark table are used to reconstruct the watermark.