Our paper titled “Fast Fallback Watermark Detection using Perceptual Hashes” was accepted for publication to the special issue “Recent Developments and Applications of Image Watermarking of MDPI Electronics. This special issue belongs to the section “Computer Science & Engineering”.
The paper proposes a fast alternative to the fallback watermarking detection system that was proposed in our previous work. Our previous work proposed to improve the robustness of existing watermarking techniques by using a so-called secondary watermark. This slow fallback system demonstrated a very high level of robustness, at the cost of a high detection complexity.
To improve our previous work, this paper proposes to speed up the detection by using small perceptual hashes instead of the large uncompressed secondary watermark signals. In this way, assuming the perceptual hashes are calculated prior to detection, the actual detection process is sped up with a factor of 26,000 to 92,000.
Using secondary watermarks improves the robustness of existing primary watermarking methods. By perceptually hashing the secondary watermarks, the fallback detection is much faster
The results presented in this study are available for download here. These results contain csv-files that contain the accuracy values (measured between the perceptual hash from all attacked watermarked videos and all compressed watermarked videos).