<p>In recent years, deep learning-based watermarking methods have been developed to address the shortcomings of traditional watermarking algorithms. Some methods adopt an end-to-end framework to train the watermarking model, enabling excellent watermark embedding and extraction. However, the visual quality and robustness of these approaches remain insufficient, especially ignoring the adequacy of image feature extraction and the correlation between network modules. We propose a feature extractor and decoder deep coupled watermark network, which can help generate high-robust watermarked images. Specifically, a down-sampling feature extractor is employed to supplement image features post-decoder, the extracted features are synchronously provided to the encoder for watermark embedding. Additionally, skip-connection is introduced to share each layer feature information of the decoder with the encoder, thereby improving the correlation between network modules. Comprehensive experimental results show that the proposed scheme can achieve high robustness against screen-shooting and paper printing processes while maintaining the visual quality of the watermarked image.</p>