Research And Embedded Implementation Of Image Watermark Algorithm

In recent years, the problem of image copyright is becoming more and more serious. With the fast development of internet and mobile communications technology, image transmission and application are universal. However, the problem of image infringement has been highlighted. Plagiarism events and fraud in photography works is nothing new. As the research and development of digital graphic technology, the technique of digital watermarking image has helped to solve this problem. In 1993, G.Garonni proposed the concept of watermarking. In 1994, R.G.Van Schyndel and some other people put forward algorithm based on LSB several years later, there appeared frequency domain digital watermarking method based on DCT, DWT and etc.The present watermarking systems are usually on the theory level, many of them are based on PC, whose shortages are high cost of system, poor stability and reliability and etc. With the development of embedded technology, embedded processor, DSP, becomes the first choice of development platform of image processing system, which takes the advantage of its small in size, high reliability, and strong function and so on. In this paper, we will study the framework of TMS320DM6437 platform of DSP C64X series , then we will study and implement the image watermarking algorithm based on LSB and DCT.LSB watermark embedding process is saving the bits of watermark image pixel values to on the lowest bits of the host image pixel values to achieve embedded watermark; DCT watermark embedding process is: the carrier image is blocked to execute DCT transform, and then we achieve embedded watermark by modifying the sub-block DC coefficient , following the key rules and the pixel values of the watermark, after that ,we implement IDCT transform on each block, finally, we obtain images with density. Watermark extraction algorithm is the inverse process of embedding algorithm.It also designed software development platform for depth study of LSB and DCT algorithms and realized it in upper computer separately and transplanted into lower computer(embedded platform DM6437).In my paper, the carrier image is the eight gray image of 256*256 BMP format, the watermark image is 32*32 binary image of BMP format. In the tests without interference, both of them are able to embed the watermark, the accuracy of watermark extraction pixel value can reach above 99%, moreover, it is running stable. However, in the test with interference, the robustness of LSB and DCT is unsatisfactory. The algorithm will be improved.