Research On Image Watermarking In Transform Domain Based On Human Visual System

The rapid development of the Internet makes digital multimedia copyright protection becoming a more and more serious problem, especially for the digital images which are widely used. New technology is needed for hiding the digital image ower’s copyright information, so the digital watermarking technology arises. Digital watermarking technology comes from book watermarking technology, which embeds the personal digital information, also known as the watermark, into a digital image and is invisible to human eye. The watermark can be extracted and checked at any time. Therefore, digital watermarking technology research has important pratical value for copyright protection.Currently there are two kinds of digital watermarking technology.One is spatial domain watermarking and the other is frequency domain watermarking. The later has become a hotspot because it has better transparency and robustness.Firstly, this paper analyzes the time-frequency characteristics of the framelet, and establishes a framelet coefficients distortion sensitive model in framelet domain based on the characteristics of human visual sensitivity. Then a one-mark-one-point non-blind watermarking algorithm is proposed in framelet domain. The visual masking value is adopted to optimize watermark embedding positions for enhancing the transparency. Moreover, then this thesis analyzes the causes of non-blind watermarking algorithm containing "plaque" effect in watermarked image and proposes a dynamic conceal method to eliminate this effect, which further enhances the transparency of the watermarking. Finally, a blind watermarking algorithm in framelet domain is proposed based on intra-prediction, in which the crowding degree is used to optimize the location of watermarking generated randomly. I This avoids the watermark bits are embedded together. In addition, segment-embedding is used and improved to optimize the embedding strength, which solve the problem of too large amplitude modification for some coefficients.. Both these two watermarking algorithms are evaluated through embedding and extracting experiments, and the experimental results show that our proposed gains better performance on transparent and robustness under common image attacks.Finally, an image watermarking software is implemented by combing C#,C++ and Matlab programming language, and the core part of image processing alogrithms are coded using matlab script...