Visual perception-based robust image watermark embedding method, watermark embedding network construction method and system, and storage medium

Abstract

The invention relates to a visual perception-based robust image watermark embedding method, a watermark embedding network construction method, a system and a storage medium, and solves the problem that the robustness and invisibility of a watermark cannot be well considered by a current watermark embedding extraction method, and the method comprises the following steps: obtaining original image information and original watermark information; extracting shallow feature image information and deep feature image information according to the original image information; analyzing the corresponding feature image information to form visual mask information; processing the original watermark information to form preprocessed watermark information; correcting the preprocessed watermark information to form watermark information to be embedded; fusing the shallow feature image information, the deep feature image information and the to-be-embedded watermark information to form fused image information; and inputting the original image information and the fused image information into the residual structure to generate final image information. According to the method, the embedding of the watermark can be regulated and controlled through the visual mask, so that the robustness and invisibility of the watermark are considered.

Description

technical field [0001] The present application relates to the technical field of image watermark embedding and extraction, in particular to a visual perception-based robust image watermark embedding method, watermark embedding network construction method, system and storage medium. Background technique [0002] Digital image watermarking technology is a kind of information hiding technology, that is, the digital watermark is embedded into the image carrier by encoding technology at the receiving end, and the complete watermark is extracted from the received watermarked image by decoding technology at the receiving end for verification. Therefore, two requirements are derived. One is to ensure that the visual quality of the carrier image is not affected when the watermark is embedded, and the other is to ensure the integrity of the watermark when extracting the watermark, that is, the invisibility and robustness of the watermark. However, the robustness and invisibility of th...

AI Technical Summary

Problems solved by technology

Among the existing deep learning methods, a CNN anti-attack network, this method uses noise anti-attack and channel coding to improve the robustness of watermarking, but this method uses channel coding error correction to increase the redundancy of embedding watermarks. The redundancy will affect the invisibility of the image watermark; in addition, a digital image steganography method with high embedding capacity is proposed based on GAN, but this method only considers the relationship between embedding capacity and image quality, and does not Robust

[0006] The above scheme has the following defects: Based on the deep learning method, the current watermark embedding extraction method cannot take into account the robustness and invisibility of the watermark, resulting in a certain room for improvement in the watermark embedding extraction process

Images