Research On Blind Digital Watermarking Algorithm Based On The Joint Time-frequency Transform

With the development of Internet and IT industry, downloading and distribution of digital media through the internet have made copyright protection of digital property a challenging problem. The digital watermarking technology, as an important branch of information hiding technology, makes the information hidden in digital images, video, audio, text documents and other digital media in order to achieve secret transmission, storage, labeling and identification, the purpose of copyright protection and tamper-resistant. Based on joint time-frequency transform theory, the blind digital watermarking algorithms are studied in this thesis which the main content are as follows:(1) Blind digital watermarking algorithm in the fractional Fourier transform domain is proposed. The original image is segmented into non-overlapping blocks for watermarking, and each block is transformed by the two dimensional fractional Fourier transform with two fractional orders. Then each pixel value of binary watermark is embedded by modifying the back-diagonal FRFT coefficients of each image block at the same location with a random array. After perform an inverse two dimensional fractional Fourier transform, we can obtain the watermarked image and the transform orders can be consider as the encryption keys in this method. A series of attacking experiments are performed on the proposed method, which show that the proposed algorithm not only is of good imperceptibility and security and is very robust to JPEG compression noise attacks and image manipulation operations, but also can provide protection even under compound attacks.(2) A Zero-watermarking method besed on the reality-perserving Fractional Fourier Transform is proposed. The original image and watermarking image are transformed by the two dimensional reality-perserving Fractional Fourier Transform with the same fractional orders. The transformed watermarking image is embedded by the zero-watermarking embedding method. Compared with other zero-watermarking algorithms, it has strong robustness against different attacks such as noise, filter and JPEG compression.