The efficient segmentation of foreground text information from the background in degraded color document images is a critical challenge in the preservation of ancient manuscripts. The imperfect preservation of ancient manuscripts over time has led to various types of degradation, such as staining, yellowing, and ink seepage, significantly affecting image binarization results. This work proposes a three-stage method using Generative Adversarial Networks (GAN) for enhancing and binarizing degraded color document images through Discrete Wavelet Transform (DWT). Stage-1 involves applying DWT and retaining the Low-Low (LL) subband images for image enhancement. In Stage-2, the original input image is divided into four single-channel images (Red, Green, Blue, and Gray), and each is trained with independent adversarial networks to extract color foreground information. In Stage-3, the output image from Stage-2 and the original input image are used to train independent adversarial networks for document binarization, enabling the integration of global and local features. The experimental results demonstrate that our proposed method outperforms other classic and state-of-the-art (SOTA) methods on the Document Image Binarization Contest (DIBCO) datasets. We have released our implementation code at https://github.com/abcpp12383/ThreeStageBinarization.

翻译：退化彩色文档图像中前景文本信息与背景的高效分割是古代手稿保存中的关键挑战。古代手稿因长期保存不当导致的各类退化（如污渍、泛黄、墨水渗透）严重影响了图像二值化效果。本文提出了一种基于生成对抗网络（GAN）和离散小波变换（DWT）的三阶段方法，用于退化彩色文档图像的增强与二值化。第一阶段应用DWT并保留低频子带图像进行图像增强；第二阶段将原始输入图像分解为四个单通道图像（红、绿、蓝、灰），分别通过独立对抗网络训练以提取彩色前景信息；第三阶段利用第二阶段输出图像与原始输入图像共同训练独立对抗网络进行文档二值化，实现全局与局部特征的融合。实验结果表明，本文方法在文档图像二值化竞赛（DIBCO）数据集上优于其他经典和最新（SOTA）方法。我们已在https://github.com/abcpp12383/ThreeStageBinarization开源实现代码。