Digital visual transmission and watermark protection of intangible cultural heritage art images based on the HAIWAN method

The art images of intangible cultural heritage undergo composite interference while being transmitted digitally. Current approaches use differentiable attack layers for adversarial attacks, but the problem is that the attack layer does not capture composite transmission distortion sufficiently. This study focuses on intangible cultural heritage images and proposed a highly applicable and highly imperceptible guided watermark attack simulation network algorithm and a Hybrid-Adaptive Image Watermark Attack Network-guided dual-domain robust blind watermarking protection model. The former continuously generates strong attack samples during training through local stability enhancement, multi-scale context aggregation, and constrained downsampling. The latter exploits the low-frequency subband in the luminance component for the wavelet transform and the frequency domain coefficients in the discrete cosine transform as dual-domain carriers, together with a channel attention embedding enhancement unit for end-to-end cooperative optimization. Experimental results show that the HAI-WAN attack samples retain a wPSNR of 35.72 dB, while their BER and NC values on three representative images range from 18.18 to 20.98% and from 75.21 to 77.97%, respectively. Compared with WatermarkAttacker and WEvade, HAI-WAN causes greater watermark extraction errors, although the remaining NC values indicate that the watermark is not completely removed. The error bit rate of the proposed model decreases to 0.0169, and the watermark detection accuracy reaches 98.68%. After introducing attack simulation training, the error bit rate under strong perturbation decreases from 15.48–29.87% to 9.86–20.74%. This technique addresses the issue of achieving a trade-off between the visual quality of art images of intangible cultural heritage and copyright marks’ robust recovery under complicated transmission situations. Not only can it protect the authenticity of intangible cultural heritage art in exhibitions, but it can also offer low latency and convenient support in media integrated communications for copyright verification.

Authors

Institutions

Publication Details

Journal
Discover Artificial Intelligence
Published
2026-10-17
DOI
https://doi.org/10.1007/s44163-026-02523-2
Primary Topic
Advanced Steganography and Watermarking Techniques
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Digital visual transmission and watermark protection of intangible cultural heritage art images based on the HAIWAN method

Yukun Zhang
Discover Artificial Intelligence
Advanced Steganography and Watermarking Techniques
article

Digital visual transmission and watermark protection of intangible cultural heritage art images based on the HAIWAN method

Yukun Zhang
article en

Abstract

The art images of intangible cultural heritage undergo composite interference while being transmitted digitally. Current approaches use differentiable attack layers for adversarial attacks, but the problem is that the attack layer does not capture composite transmission distortion sufficiently. This study focuses on intangible cultural heritage images and proposed a highly applicable and highly imperceptible guided watermark attack simulation network algorithm and a Hybrid-Adaptive Image Watermark Attack Network-guided dual-domain robust blind watermarking protection model. The former continuously generates strong attack samples during training through local stability enhancement, multi-scale context aggregation, and constrained downsampling. The latter exploits the low-frequency subband in the luminance component for the wavelet transform and the frequency domain coefficients in the discrete cosine transform as dual-domain carriers, together with a channel attention embedding enhancement unit for end-to-end cooperative optimization. Experimental results show that the HAI-WAN attack samples retain a wPSNR of 35.72 dB, while their BER and NC values on three representative images range from 18.18 to 20.98% and from 75.21 to 77.97%, respectively. Compared with WatermarkAttacker and WEvade, HAI-WAN causes greater watermark extraction errors, although the remaining NC values indicate that the watermark is not completely removed. The error bit rate of the proposed model decreases to 0.0169, and the watermark detection accuracy reaches 98.68%. After introducing attack simulation training, the error bit rate under strong perturbation decreases from 15.48–29.87% to 9.86–20.74%. This technique addresses the issue of achieving a trade-off between the visual quality of art images of intangible cultural heritage and copyright marks’ robust recovery under complicated transmission situations. Not only can it protect the authenticity of intangible cultural heritage art in exhibitions, but it can also offer low latency and convenient support in media integrated communications for copyright verification.

Discover Artificial IntelligenceVol. 6(1)
Weinan Normal University (CN)
Weinan Normal University
Openalex Percentile: Top 16%
Advanced Steganography and Watermarking Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.