| Summary: | 碩士 === 國立臺中技術學院 === 資訊科技與應用研究所 === 95 === Copyright protection, data integrity and confidential communication have been paid much more attention in recent years, and avoiding torts or counterfeits means that the loss is reduced. In today’s digital era, the chance of torts of digital images increases because digital images are easy to be copied, modified and transmitted. Information hiding is an efficient method for copyright protection, data integrity and confidential communication. The widely used techniques are robust watermarking techniques, fragile watermarking techniques and steganography according to its purpose. These techniques embed the information into digital images, but their purposes are different. The requirement of robust watermarking is that the watermark received from the watermarked image can be recognized after different kinds of attacks. It is used for copyright protection. The purpose of fragile watermarking is to determine whether the image is tampered or not and to locate the tampered region. It is used to protect image integrity. Steganography uses an image to be a camouflage to transmit the secret information, and avoids illegal users discovering the secret information.
This thesis presents three information hiding schemes in the transformation domain according to the image characteristics, including a robust watermarking, a semi-fragile watermarking and a reversible data hiding. The coefficients in the transformation domain can represent image characteristics. The robustness of the coefficients in different subbands is different, and modifying them will cause different distortions. Embedding the information into the coefficients in different subbands can effectively adjust the robustness and the image quality.
In Chapter 3, a robust watermarking scheme resisting both compression attacks and geometrical attacks is proposed. An image is transformed into discrete wavelet coefficients by using discrete wavelet transformation, and two of the coefficients having the same indices in HL subband and LH subband are used to form a gradient vector. The direction of the gradient vector is adjusted to embed the watermark, and the length is increased to improve the robustness of the watermark.
In Chapter 4, a semi-fragile watermarking scheme based on singular value decomposition (SVD) and Vector Quantization (VQ) is proposed. The SVD coefficients of U matrix and V matrix are processed by VQ to extract the image features, and these features are embedded into the biggest singular value of each image sub-block. The main idea behind the proposed scheme is to embed image features obtained from SVD coefficients, into the protected image. If image features or the embedded information are modified, the tampered regions can be detected.
In Chapter 5, a reversible data hiding scheme is proposed. Based on the integer wavelet transformation, the secret data is embedded into the transformation domain, and the stego-image can be recovered to the original image losslessly. Because the modification in the high frequency subbands is not sensitive to human vision, the image quality can be improved. The proposed scheme uses the coefficient expansion to embed the secret data. The coefficients of three 2-level high frequency subbands are used to predict the absolute value of the coefficients of three 1-level high frequency subbands. From the prediction, the small absolute value of the coefficients of three 1-level high frequency subbands can be determined to embed one secret bit or two secret bits. This scheme can increase the capacity effectively.
|
|---|
