Research On Image Digital Watermarking Technology Based On CUDA

With the continuous development of the computer-science, digital media such as image, audio and video have occupied people's lives. The popular of the digital content makes the security and the legality of digital media to be the much concerned issue. Digital watermarking technology has developed rapidly in recent years, has became the important guarantee in digital product's copyright protection terms. With the continuous improvement of the watermark crack technology, to achieve the requirements of higher quality of the image effect and more secure of the information protection, digital watermarking technology gradually to be the development of large-scale and complex. With the emerging of new electronic products, watermarking products has been produced by the scale of hundreds of millions each year. The time delay of digital watermarking algorithm in high complex situation has been continuously enlarged, not only wasted a lot of resources, but also brought the risk to the watermarking information. CUDA compute unified device architecture produced by NVIDIA, according to structural features of graphics hardware, taking the parallel instruction set as the operation mode, has rapid development in recent years and used in a large number of scientific computing optimization algorithm. The paper take the using model of the GPU large-scale parallel computing as the exploration target, start from the traditional problems, choose the typical algorithms to study, mainly research the parallel of the image digital watermarking algorithms, the time domain, the frequency domain and the compression domain on GPU. The traditional image watermarking has the features of tedious iteration and not suitable for parallel computing. The paper proposes the formula transformation strategy based on the transformation matrix, solve the parallel computational problems of traditional watermarking embedding algorithm problems by matrix calculating. On the basis of the watermark parallel computing, the paper further study the optimization strategy of the parallel computing, including the shared memory and the stream compute on GPU, further improve the efficiency of the watermarking algorithms. By the serial and parallel image digital watermarking technology comparison test, the paper proved that the image digital watermarking technology based on CUDA greatly improved the efficiency of the efficiency without lower the image quality, the image robust and the watermark hidden.