Research On VLSI-based Digital Image Watermarking Technology

Digital image watermarking plays an important role in image copyright protection and the authenticity certification. In the past few decades, researchers have proposed a large number of software-based image watermarking schemes. Now software-based image watermarking algorithms have attained a certain degree of maturity, and VLSI-based image watermarking algorithm is in their infancy. The advantages of VLSI-based watermarking are low cost, good real-time performance, high reliability, and ease of integration with existing consumer electronic devices. Therefore, hardware-base watermarking technology has important practical value.After deeply study the digital watermarking principle, the dissertation focuses on the research of copyright protection watermarking algorithm, multi-purpose watermarking algorithm , and copyright marking watermarking algorithm, and concerned VLSI architecture design, implementation and optimization. The application of Human Visual System in copyright protection watermarking technology, the affection of JPEG compression to the relationship of neighboring DCT and DWT coefficients and the application of Human Eye Gradation Sensitive Characteristics in visible watermarking technology are detailed explored in this dissertation. Via the attack tests, the robustness of the three watermarking algorithms is analyzed in detail. The experiment results indicate that both the copyright protection watermarking algorithm and multipurpose watermarking algorithm have good invisibility and robustness, the Peak Signal to Noise Ratio of the watermarked images are all above 41dB, the Normalized Coefficient values of extracted watermarks are all above 0.95. The visible watermarking algorithm has good visibility and the robustness. Finally, the algorithms are integrated in Image Signal Processor (ISP) system, and verificated via Altera Cyclone FPGA. The experiment result indicates that the hardware implementation of the three digital watermarking algorithms can be well integrated in ISP system, and can meet the requirement of real-time image processing. All three algorithms'maximum operating frequencies are all above 120MHz. Therein, the copyright protection blind watermarking algorithm is implemented via ASIC, using Chartered 0.18μm CMOS technology, and has been taped out.The creative works of this dissertation include: proposing an adaptively blind watermarking technology for image copyright protection. According to the characteristics of Human Visual System (HVS), watermark stretch factor is adaptively chosen through analyzing the texture intensity of the image. Blind watermarking embedding is achieved by averaging the neighboring DCT coefficients in middle frequency. Compared with other copyright protection watermarking algorithms in DCT domain, it has the advantage of simple architecture and ease of VLSI implementation. Proposing a multipurpose blind watermarking technology with resistance to JPEG compression. Based on analyzing the relationship of neighboring DWT coefficients before and after the JPEG compression, the technology embeds the robust watermark and semi-fragile watermark into different DWT sub-band. The technology has a good performance against JPEG compression, and can accurately detect malicious attacks. Compared with other multipurpose watermarking algorithms, it has the advantage of simple computation and higher Peak Signal to Noise Ratio value of watermarked image. Proposing a visible watermarking technology based on Bayer image. Via setting up the function of Human Eye Gradation Sensitive Characteristics, it embeds visible watermark in Bayer data before interpolation. Compared with other visible watermarking algorithms, it ensures the consistency of visible watermark in different image brightness areas. The method is computationally simple, and easy to implement.