| Popis: |
Protecting ownership of digital content is challenging in today’s online world, especially when sharing content through social networks and instant messengers. One possible solution is the use of watermarking; however, if the watermarking method is not robust enough, the watermark can get damaged or erased during transmission. This study introduces a template-based watermarking method with neural network-based extraction, strengthened by error-correcting polar codes, that is designed to work well on major social networks like Facebook, Facebook Messenger, VK, Telegram, Snapchat, Pinterest, and WhatsApp. The method can embed imperceptible (PSNR=39.66) watermarks ranging from 650 to 1600 bits for high-definition images (1920-by-1080 pixels) with small probability (0.01) of erroneous extraction after being transferred through social networks. The key feature of the method is its ability to work with real communication channels, as shown during its testing. |
