The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XLIV-2/W1-2021
https://doi.org/10.5194/isprs-archives-XLIV-2-W1-2021-47-2021
https://doi.org/10.5194/isprs-archives-XLIV-2-W1-2021-47-2021
15 Apr 2021
 | 15 Apr 2021

PATCH-BASED ADAPTIVE IMAGE AND VIDEO WATERMARKING SCHEMES ON HANDHELD MOBILE DEVICES

M. N. Favorskaya and V. V. Buryachenko

Keywords: Video Watermarking Scheme, Patch-based Search, Relevant Region, Static Scene, Mobile Device

Abstract. Mobile devices provide a huge amount of multimedia information sending to the members of social groups every day. Sometimes it is required to authorize the sending information using the limited computational resources of smartphones, tablets or laptops. The hardest problem is with smartphones, which have the limited daily energy and battery life. There are two scenarios for using mobile watermarking techniques. The first scenario is to implement the embedding and extraction schemes using proxy server. In this case, the watermarking scheme does not differ from conventional techniques, including the advanced ones based on adaptive paradigms, deep learning, multi-level protection, and so on. The main issue is to hide the embedding and extracting information from the proxy server. The second scenario is to provide a pseudo-optimized algorithm respect to robustness, imperceptibility and capacity using limited mobile resources. In this paper, we develop the second approach as a light version of adaptive image and video watermarking schemes. We propose a simple approach for creating a patch-based set for watermark embedding using texture estimates in still images and texture/motion estimates in frames that are highly likely to be I-frames in MPEG notation. We embed one or more watermarks using relevant large-sized patches according to two main criteria: high texturing in still images and high texturing/non-significant motion in videos. The experimental results confirm the robustness of our approach with minimal computational costs.