Watermark Image Reconstruction Based on Deep Learning
Qing Xin Yang, Yiming Zhang, Long Wang, Wenbo Zhao · 2019 International Conference on Sensing, Diagnostics, Prognostics, and Control (SDPC) · 2019
To solve the problem of low watermark reading rate caused by external factors when trademark images identify watermark information, this paper proposes a watermark image reconstruction model based on super resolution and deep learning. The model consists of two parts: feature extraction and image reconstruction. Seven convolution layers with 3*3 convolution kernels and skip connections form the feature extraction part. Several 1*1 convolution layers form the image reconstruction part. The watermark reading rate is evaluated by the number of watermark blocks read and the average watermark score. Experimental results show that the algorithm has better performance for watermark image reconstruction.