Jiaohua Qin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Jiaohua Qin has a total of 108 co-authors. Here are the top co-authors by number of shared publications:
- Xuyu Xiang
Central South University of Forestry and Technology, Changsha, China
37 shared publications - Yun Tan
Central South University of Forestry and Technology, Changsha, China
36 shared publications Xuyu Xiang
33 shared publicationsYun Tan
21 shared publications- Yuanjing Luo
Central South University of Forestry and Technology, Changsha, Hunan, China
8 shared publications - Naixue Xiong
Sul Ross State University
7 shared publications
Latest Publications
- 19 October 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 334
Image-semantic guided text steganography based on LLMs and affine feature fusion
- 19 October 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 334
Image-semantic guided text steganography based on LLMs and affine feature fusion
- 10 August 2026
- Published by Elsevier in Knowledge-Based Systems
- Vol. 351
Reverse verification-based model watermarking with resistance to model extraction attacks
- 10 August 2026
- Published by Elsevier in Knowledge-Based Systems
- Vol. 351
Reverse verification-based model watermarking with resistance to model extraction attacks
- 21 July 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 332
Dual-domain multi-scale and edge-guided network for image manipulation localization
- 21 July 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 332
Dual-domain multi-scale and edge-guided network for image manipulation localization
- 18 May 2026
- Published by Elsevier in Pattern Recognition
- Vol. 180
SSRW-INN: An invertible neural network for screen shooting robust watermarking
- 18 May 2026
- Published by Elsevier in Pattern Recognition
- Vol. 180
SSRW-INN: An invertible neural network for screen shooting robust watermarking
- 20 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Video Technology
- Vol. PP (10518215) , 1
DCL-Net: Decoupled Contrastive Learning Network for Image Manipulation Localization
Extracting discriminative forensic artifacts from high-dimensional latent spaces is pivotal for Image Manipulation Localization (IML). Nevertheless, the intrinsic feature entanglement arising from subtle structural discrepancies continues to impede precise localization. Specifically, manipulation artifacts are frequently overwhelmed by coherent background textures or ’soft boundaries’, rendering the trace-rich features inextricably mixed with intrinsic image content. To disentangle these intertwined representations, we propose the Decoupled Contrastive Learning Network (DCL-Net) for robust image tampering localization. Leveraging a Vision Transformer (ViT) backbone integrated with a Simple Feature Pyramid Network (SFPN), DCL-Net incorporates a novel Feature Decoupling Module (FDM). The FDM explicitly disentangles the feature space into foreground, background, and uncertainty regions-thereby effectively isolating manipulation cues from coherent background textures and capturing the transitional nature of ambiguous boundaries. Furthermore, to align the optimization objective with the intrinsic structure of manipulation traces, we introduce a prior-guided contrastive learning strategy that explicitly ’pushes away’ manipulated features from authentic and uncertain components. Finally, the Contrast Feature Aggregation Module (CFAM) employs a two-stage attention mechanism to refine these disentangled features, further suppressing redundant background details. Extensive experiments across five public benchmarks demonstrate that DCL-Net delivers state-of-the-art localization performance and exhibits strong robustness against common distortions.
- 20 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Video Technology
- Vol. PP (10518215) , 1
DCL-Net: Decoupled Contrastive Learning Network for Image Manipulation Localization
Extracting discriminative forensic artifacts from high-dimensional latent spaces is pivotal for Image Manipulation Localization (IML). Nevertheless, the intrinsic feature entanglement arising from subtle structural discrepancies continues to impede precise localization. Specifically, manipulation artifacts are frequently overwhelmed by coherent background textures or ’soft boundaries’, rendering the trace-rich features inextricably mixed with intrinsic image content. To disentangle these intertwined representations, we propose the Decoupled Contrastive Learning Network (DCL-Net) for robust image tampering localization. Leveraging a Vision Transformer (ViT) backbone integrated with a Simple Feature Pyramid Network (SFPN), DCL-Net incorporates a novel Feature Decoupling Module (FDM). The FDM explicitly disentangles the feature space into foreground, background, and uncertainty regions-thereby effectively isolating manipulation cues from coherent background textures and capturing the transitional nature of ambiguous boundaries. Furthermore, to align the optimization objective with the intrinsic structure of manipulation traces, we introduce a prior-guided contrastive learning strategy that explicitly ’pushes away’ manipulated features from authentic and uncertain components. Finally, the Contrast Feature Aggregation Module (CFAM) employs a two-stage attention mechanism to refine these disentangled features, further suppressing redundant background details. Extensive experiments across five public benchmarks demonstrate that DCL-Net delivers state-of-the-art localization performance and exhibits strong robustness against common distortions.
- 6 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 64 (01962892) , 1-15
A Dual-Branch Collaborative Network Based on CNN and Transformer With Multiscale Dual Coordinate Attention for Farmland Segmentation
Accurate farmland segmentation from satellite imagery is crucial for precision agriculture and environmental monitoring, allowing optimal resource allocation and sustainable agricultural development. Current farmland segmentation methods rely primarily on convolutional neural networks (CNNs) or self-attention mechanisms (Transformers) for feature modeling. However, CNNs are limited by local receptive fields, struggling to capture global context, whereas Transformers face challenges with quadratic computational complexity. Although some CNN–Transformer hybrid approaches aim to integrate complementary local and global information to improve performance, they do not adequately address the boundary confusion problem caused by narrow interfield structures, such as ridges and roads between agricultural plots. To this end, we propose a dual-branch collaborative network DCCLNet to tackle the challenge of separating the boundaries of farmland. We design a multiscale dual coordinate attention (MDCA) that captures boundary-oriented features along height and width dimensions through strip-pooling strategies with adaptive weighting, specifically targeting narrow boundary structures. We propose a rectified logitwise collaborative learning (RLCL) strategy that integrates multiscale boundary detection with adaptive reliability estimation, performing enhanced correction in boundary regions through dynamic weight allocation. We propose a class-aware featurewise collaborative learning (CFCL) strategy that employs hierarchical prototype-based category perception (HPCP) at multiple spatial pyramid levels. This achieves effective feature-level collaboration between the CNN and Transformer branches in diverse field sizes. Extensive experiments on three farmland datasets demonstrate that DCCLNet outperforms most state-of-the-art methods in multiple evaluation metrics, including intersection over union (IoU), ${F}1$ -score, precision, and recall.
- 6 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 64 (01962892) , 1-15
A Dual-Branch Collaborative Network Based on CNN and Transformer With Multiscale Dual Coordinate Attention for Farmland Segmentation
Accurate farmland segmentation from satellite imagery is crucial for precision agriculture and environmental monitoring, allowing optimal resource allocation and sustainable agricultural development. Current farmland segmentation methods rely primarily on convolutional neural networks (CNNs) or self-attention mechanisms (Transformers) for feature modeling. However, CNNs are limited by local receptive fields, struggling to capture global context, whereas Transformers face challenges with quadratic computational complexity. Although some CNN–Transformer hybrid approaches aim to integrate complementary local and global information to improve performance, they do not adequately address the boundary confusion problem caused by narrow interfield structures, such as ridges and roads between agricultural plots. To this end, we propose a dual-branch collaborative network DCCLNet to tackle the challenge of separating the boundaries of farmland. We design a multiscale dual coordinate attention (MDCA) that captures boundary-oriented features along height and width dimensions through strip-pooling strategies with adaptive weighting, specifically targeting narrow boundary structures. We propose a rectified logitwise collaborative learning (RLCL) strategy that integrates multiscale boundary detection with adaptive reliability estimation, performing enhanced correction in boundary regions through dynamic weight allocation. We propose a class-aware featurewise collaborative learning (CFCL) strategy that employs hierarchical prototype-based category perception (HPCP) at multiple spatial pyramid levels. This achieves effective feature-level collaboration between the CNN and Transformer branches in diverse field sizes. Extensive experiments on three farmland datasets demonstrate that DCCLNet outperforms most state-of-the-art methods in multiple evaluation metrics, including intersection over union (IoU), ${F}1$ -score, precision, and recall.
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28 (15209210) , 2643-2653
PPDSA: Privacy-Preserving Cross-Modal Retrieval With Disentangled Soft-Label Alignment
Cross-modal retrieval techniques, with their efficient search capabilities, have garnered considerable attention in both industry and academia. Among them, image-text cross-modal retrieval, a classic task in the cross-modal domain, has been widely applied in internet applications. However, image data in cross-modal retrieval applications often contains rich personal sensitive information, which is typically stored in plaintext format, posing a risk of privacy leakage. Due to the unique semantic and heterogeneous gap in the cross-modal space, existing single-modality image privacy protection techniques cannot be directly applied to cross-modal scenarios. To address this issue, this paper proposes a privacy-preserving cross-modal retrieval method based on disentangled soft-label alignment(PPDSA). Firstly, to protect the privacy of image data, thumbnail preserving encryption is used in this paper. This method aims to encrypt the image while ensuring the feasibility of graph-text cross-modal feature representation learning, so as to realize cross-modal encrypted retrieval. Second, to further improve the retrieval effect of encrypted images and normal text, this paper introduces a disentangled soft label alignment technique. Specifically, a teacher model is used to obtain the soft label matrix, and cross-modal (encrypted image-text pair) and single-modal (encrypted image pair) soft label alignment techniques are designed to capture more fine-grained semantic representation information and reduce the interference of false positives on similarity recognition. A series of experiments on Flickr30k and MSCOCO datasets show that the proposed method not only effectively protects image privacy in cross-modal retrieval, but also improves the retrieval performance by 6.7% and 3.7% respectively compared with the CLIP baseline.
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28 (15209210) , 2643-2653
PPDSA: Privacy-Preserving Cross-Modal Retrieval With Disentangled Soft-Label Alignment
Cross-modal retrieval techniques, with their efficient search capabilities, have garnered considerable attention in both industry and academia. Among them, image-text cross-modal retrieval, a classic task in the cross-modal domain, has been widely applied in internet applications. However, image data in cross-modal retrieval applications often contains rich personal sensitive information, which is typically stored in plaintext format, posing a risk of privacy leakage. Due to the unique semantic and heterogeneous gap in the cross-modal space, existing single-modality image privacy protection techniques cannot be directly applied to cross-modal scenarios. To address this issue, this paper proposes a privacy-preserving cross-modal retrieval method based on disentangled soft-label alignment(PPDSA). Firstly, to protect the privacy of image data, thumbnail preserving encryption is used in this paper. This method aims to encrypt the image while ensuring the feasibility of graph-text cross-modal feature representation learning, so as to realize cross-modal encrypted retrieval. Second, to further improve the retrieval effect of encrypted images and normal text, this paper introduces a disentangled soft label alignment technique. Specifically, a teacher model is used to obtain the soft label matrix, and cross-modal (encrypted image-text pair) and single-modal (encrypted image pair) soft label alignment techniques are designed to capture more fine-grained semantic representation information and reduce the interference of false positives on similarity recognition. A series of experiments on Flickr30k and MSCOCO datasets show that the proposed method not only effectively protects image privacy in cross-modal retrieval, but also improves the retrieval performance by 6.7% and 3.7% respectively compared with the CLIP baseline.