总结总结一下,本文提出了一种新的水印去除方法,相比现有方法在彩色水印数据集CLWD上有可观的提升,并且具备了自纠正的水印检测功能,大大提高了水印检测的质量,同时也帮助了后续的水印区域的图像修复工作;我们还提出了跨阶段与跨层次的信息融合背景修复模块,有效地提高了图片的修复质量。此外通过实验,我们发现,纯颜色的水印比较容易被现有的方法检测并擦除,若是比较复杂的颜色、纹理的图案,特别是有细碎的图案环绕在主图案的时候,现有的模型处理起来比较困难,因而水印的设计应考虑更多这些细节。水印去除方向已有的论文和代码已经总结在https://github.com/bcmi/Awesome-Visible-Watermark-Removal。参考文献:[1] Cao, Zhiyi, et al. "Generative adversarial networks model for visible watermark removal."IET Image Processing13.10 (2019): 1783-1789.[2] Li, Xiang, et al. "Towards photo-realistic visible watermark removal with conditional generative adversarial networks."International Conference on Image and Graphics, 2019.[3] Hertz, Amir, et al. "Blind visual motif removal from a single image."Proceedings of CVPR, 2019.[4] Liu, Yang, Zhen Zhu, and Xiang Bai. "WDNet: Watermark-Decomposition Network for Visible Watermark Removal." WACV, 2021.[5] Cheng, Danni, et al. "Large-scale visible watermark detection and removal with deep convolutional networks."PRCV, 2018.