Anti-screenshot Watermarking Algorithm About Archives Image Based on Deep Learning Model

Published: 01 Jan 2022, Last Modified: 13 Nov 2024ICS 2022EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: In recent years, there are an increasing number of incidents in which archives images have been ripped. Leak tracking is possible by adding an anti-screenshot digital watermark to an archive image. However, because an archives image's texture is single, there is a problem of low detection rate of watermark with the existing algorithm. So in order to improve the robustness of archives image anti-screenshot, we propose an anti-screenshot deep learning model (DLM): ScreenNet. It aims to enhance the background and enrich the texture with style transfer. Firstly, a preprocessing process based on style transfer is added before the archives image is input into the encoder. Secondly, the ripped images usually have moiré, so we generate a database of ripped archives images with moiré by means of a moiré network. Lastly, by improving the Stagstamp model, the watermark information is encoded/decoded through the improved ScreenNet model using the ripped archives image database as the noise layer. The experiment proves that the algorithm is able to resist anti-screenshot attacks and achieve the ability to detect watermark information to leak trace of ripped images.
Loading