Robust discrete cosine transform based technique for image watermarking against cropping attacks

• Main Author: Rajabi, Muhammad Javad
• Format: Thesis
• Language: English
• Published: 2018
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://eprints.utm.my/id/eprint/96253/1/MuhammadJavadRajabiPAIS2018.pdf.pdf; http://eprints.utm.my/id/eprint/96253/; http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:142874

Watermarking is a common technique for authentication and message hiding. Watermarking should be invisible and robust to common processing and attack. So far, current research has succeeded in maintaining the high normalized correlation (NC) quality of watermark only up to 25 percent cropping level. The purpose of this research is to enhance the current quality of NC and maintain the robustness of watermark in higher cropping level. This research was divided into four phases. First, this project analyzed different frequency domain watermarking techniques against cropping attacks. Second, a discrete cosine transform (DCT) based technique for image watermarking against cropping attacks was synthesized from the literature. The proposed watermarking technique hides watermark image as logo into a host image. The host image was first divided into 8x8 blocks, and then DCT transformation was applied on each block. Next, Arnold’s cat map was applied on watermark image and embedded into the host image using zigzag symmetric technique introduced in this research. In the third phase, the design of improved watermarking technique was tested and evaluated against cropping attack. Experiment results showed that the proposed algorithm was undetectable and robust against 50% cropping attack. Several kinds of cropping attacks on the watermarked image were implemented, which included top half cropping, right half cropping, bottom half cropping, and left half cropping. The watermark from attacked watermarked image was then extracted and compared to the original watermark logo using NC. Even though the watermarked image was impaired, the watermark remained almost intact with high quality of NC. The compression of the proposed method showed a decrease in size of the image compare with JPEG compression. The results of NC for extracted watermark from five standard watermarked Lena, Peppers, Baboon, Goldhill and Barbara images within acceptable value of NC, that was above 0.99 after 50% of cropping attack. In addition, compression comparison between the proposed method and other watermarking techniques based on the Lena image as the host image showed the results surpassed 5% to 25% of the current results in the literature. Therefore, this research has improved the robustness of image watermarking against cropping and compression attacks.