23 September 2025 in IEEE Transactions on Circuits and Systems for Video Technology
Previous INN-based image hiding methods have demonstrated outstanding concealing and revealing performances; however, their robustness is compromised due to the unbiased and lossless characteristics inherent in strictly symmetric INNs. To address this limitation, we propose the Degradation-Adaptive Semi-Symmetric Network (DASS-Net) for robust image hiding. DASS-Net incorporates asymmetric units into INN flow, injecting nonlinearity and adaptability to the bijection. Specifically, Coupled Learning Unit (CLU) is proposed as asymmetric unit in the concealing phase, to facilitate more flexible information interaction. While the revealing process incorporates Degradation-Aware Modulation Units (DAMU) as asymmetric units that capture degradation clues and customize modulation parameters, thus enabling dynamic adjustments for optimal revealing under various degradation scenarios. Additionally, we propose a Frequency Sub-bands Enhancement Module (FSEM) that leverages low-frequency features to assist the recovery of high-frequency features and strengthen hidden information, thereby mitigating information loss caused by degradation. Extensive experiments on COCO and DIV2K datasets, exhibiting a range of degradation types and degrees, indicate that DASS-Net markedly surpasses state-of-the-art methods, with more than 10% improvements in imperceptibility and robustness.
1 September 2026 in IEEE Internet of Things Journal
22 May 2026 in IEEE Transactions on Multimedia
Image hiding is a prevalent technique for transmitting secret images in an open environment. Previous Invertible Neural Network (INN)-based approaches exhibit outstanding hiding capabilities. However, they struggle to simultaneously improve imperceptibility and robustness in distorted scenarios, attributable to the inherently limited nonlinearity and strictly reversible characteristics of INNs. To counteract this problem, we present a novel framework named Degradation Guided Robust Image Hiding. This framework integrates heterogeneous units into INN flow, injecting nonlinearity and adaptability to concealing and revealing processes. Specifically, cross learning unit (CLU) is proposed in the concealing phase, which enhances nonlinear capability by exploring more flexible information exchange mechanism and hiding areas. The revealing network incorporates degradation guided units (DGU) that capture degradation representations based on contrastive learning, and then customize modulation parameters to enable dynamic adjustments of revealing behaviors according to various distortions. Comprehensive experiments conducted on COCO and DIV2K datasets, which feature a spectrum of distortions, demonstrate that our method significantly outperforms state-of-the-art methods, with improvements in imperceptibility exceeding 4.2 dB and enhancements in robustness exceeding 3.1 dB across six scenarios. Experimental results show that our method not only enhances the robustness in various degraded scenarios but also improves imperceptibility.