1. 主页
  2. 科研成果

科研成果

年

2025

  1. local uncertainty energy transfer for active domain adaptation

    Y. Sun, G. Shi, W. Dong, X. Li, L. Dong, and X. Xie
    IEEE Trans. on Image Processing, 2024 (TIP)
     [Paper] [Project Page] [code] [demo]

  2. CDS-Net: Contextual Difference Sensitivity Network for Pixel-Wise Road Crack Detection
    We propose a novel method in this paper, CDS-Net, that significantly improves road crack detection performance through multiple practical modules, including the Multi-Directional Hierarchical Attention (MDHA) module and the Difference Sensitivity Reconstruction Block (DSRB).
    Q. Tan, A. Li, L. Dong, W. Dong, X. Li and G. Shi
    IEEE Transactions on Circuits and Systems for Video Technology, 2024 (T-CSVT)
     [Paper] [Project Page] [code] [demo]


  3. Bridging Task Boundaries: Remote Sensing ImageText Retrieval via Dictionary-Driven Adaptation
    Junwei Xu, Tao Huang, Zhenyu Wang, Weisheng Dong, and Xin Li
    IEEE International Conference on Acoustics, Speech, and Signal Processing, 2025. (ICASSP)
     [Paper] [Project Page] [code] [demo]

2024

  1. Spatial-Spectral Mixing Transformer With Hybrid Image Prior for Multispectral Image Demosaicing
    We propose a novel MSI demosaicing method based on the spatial-spectral mixing transformer with hybrid image prior, named SSMT-HIP, to enhance image reconstruction and detail recovery.
    L. Dong, M. Liu, T. Tang, T. Huang, J. Lin, W. Dong, and G. Shi
    IEEE Journal of Selected Topics in Signal Processing, 2024 (J-STSP)
     [Paper] [Project Page] [code] [demo]

  2. Discriminative Correspondence Estimation for Unsupervised RGB-D Point Cloud Registration
    We design a generative feature extraction module to fully leverage multimodal information, and seek a novel perspective for correspondence estimation which expands the points in the source and target point clouds into hyperrectangle-based embeddings and considers their inner relationships .
    Chenbo Yan, Mingtao Feng, Zijie Wu, Yulan Guo, Weisheng Dong, Yaonan Wang and Ajmal Mian
    IEEE Trans. on Circuits and Systems for Video Technology, 2024 (TCSVT)
     [Paper] [Project Page] [code] [demo]

  3. Learning real-world heterogeneous noise models with a benchmark dataset
    We first construct a more comprehensive dataset of the real world by capturing more indoor and outdoor scenes under different lighting conditions using diverse smartphones, then we propose a non-parametric noise estimation method capable of modeling the spatial heterogeneity of real-world noise patterns .
    L. Sun, J. Lin, W. Dong, X. Li, J. Wu, and G. Shi
    Pattern Recognition, in press (PR)
     [Paper] [Project Page] [code] [demo]

  4. Multi-scale spatio-temporal memory network for lightweight video denoising
    We exploit a multiscale representation based on the Gaussian-Laplacian pyramid decomposition .
    Lu Sun, Fangfang Wu, Wei Ding, Xin Li, Weisheng Dong, and Guangming Shi
    IEEE Trans. on Image Processing, in press, 2024 (TIP)
     [Paper] [Project Page] [code] [demo]

  5. TSUDepth: exploring temporal symmetry-based uncertainty for unsupervised monocular depth estimation
    We introduce an innovative framework known as Temporal Symmetry-based Uncertainty (TSU)-Depth, aimed at enhancing the accuracy of unsupervised monocular depth estimation.
    Y. Zhu, R. Ren, W. Dong, X. Li and G. Shi
    Neurocomputing, 2024 (IJON)
     [Paper] [Project Page] [code] [demo]

  6. External Knowledge Enhanced 3D Scene Generation from Sketch
    We propose a sketch based knowledge enhanced diffusion architecture (SEK) for generating customized, diverse, and plausible 3D scenes.
    Zijie Wu, Mingtao Feng, Yaonan Wang, He Xie, Weisheng Dong, Bo Miao, and Ajmal Mian
    ECCV 2024 (ECCV)
     [Paper] [Project Page] [code] [demo]

  7. Uncertainty modeling of the transmission map for single image dehazing
    We propose to model the uncertainty in the estimation of the transmission map and develop a spatially adaptive learning module for ASM correction.
    Bokang Wang, Qian Ning, Fangfang Wu, Xin Li
    IEEE Trans. on Circuits and Systems for Video Technology, 2024 (TCSVT)
     [Paper] [Project Page] [code] [demo]

  8. Inverse weight-balancing for deep long-tailed learning
    We propose an inverse weight-balancing (IWB) approach to guide model training and alleviate the data imbalance problem in two stages.
    Wenqi Dang, Zhou Yang, Weisheng Dong, Xin Li
    Association for the Advancement of Artificial Intelligence, 2024 (AAAI)
     [Paper] [Project Page] [code] [demo]

  9. TransVQA: transferable vector quantization alignment for unsupervised domain adaption
    We present a novel model named Transferable Vector Quantization Alignment for Unsupervised Domain Adaptation (TransVQA), which integrates the Transferable transformer-based feature extractor (Trans), vector quantization domain alignment (VQA), and mutual information weighted maximization confusion matrix (MIMC) of intra-class discrimination into a unified domain adaptation framework.
    Yulin Sun, Weisheng Dong, Xin Li, Le Dong, Guangming Shi, and Xuemei Xie
    IEEE Trans. on Image Processing, 2024. (TIP)
     [Paper] [Project Page] [code] [demo]

  10. BVT-IMA: Binary Vision Transformer with Information-Modifed Attention
    We find a correlation between attention scores and the information quantity, further indicating that a reason for such a phenomenon may be the loss of the information quantity induced by constant moduli of binarized tokens.
    Zhenyu Wang, Hao Luo, Xuemei Xie, Fan Wang, Guangming Shi
    Association for the Advancement of Artificial Intelligence, 2024 (AAAI)
     [Paper] [Project Page] [code] [demo]

2023

  1. Exploring Correlations in Degraded Spatial Identity Features for Blind Face Restoration
    We propose a novel method that explores the correlation of degraded spatial identity features by learning a general representation using memory network.
    Q. Ning, F. Wu*, W. Dong, X. Li, and G. Shi
    ACM Multimedia, 2023 (ACMMM)
     [Paper] [Project Page] [code] [demo]

  2. Low-Light image enhancement with multi-stage residue quantization and brightness-aware attention
    We propose a brightness-aware network with normal-light priors based on brightness-aware attention and residual quantized codebook.
    Y. Liu, T. Huang, W. Dong*, X. Li, and G. Shi
    IEEE ICCV, 2023 (ICCV)
     [Paper] [Project Page] [code] [demo]

  3. Spatially Varying Prior Learning for Blind Hyperspectral Image Fusion
    we propose a deep blind HIF method by unfolding model-based maximum a posterior (MAP) estimation into a network implementation.
    J. Xu, F. Wu, X. Li, W. Dong, T. Huang, and G. Shi
    IEEE Trans. on Image Processing, 2023 (TIP)
     [Paper] [Project Page] [code] [demo]

  4. Uncertainty-Driven Knowledge Distillation for Language Model Compression
    We propose a novel and efficient uncertainty-driven knowledge distillation compression method for transformer-based pretrained language model.
    T. Huang, W. Dong*, F. Wu, X. Li, and G. Shi
    IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2023 (TASLP)
     [Paper] [Project Page] [code] [demo]

  5. Memory based temporal fusion network for video deblurring
    We proposed a memory-based temporal fusion network (TFN) to capture local spatial-temporal relationships across the input sequence for video deblurring.
    C. Wang, W. Dong*, X. Li, F. Wu, J. Wu, and G. Shi
    International Journal of Computer Vision, 2023 (IJCV)
     [Paper] [Project Page] [code] [demo]

  6. Deep Gaussian Scale Mixture Prior for Image Reconstruction
    This paper proposes a novel image reconstruction method based on the Maximum a Posterior (MAP) estimation framework using learned Gaussian Scale Mixture (GSM) prior.
    T. Huang, X. Yuan, W. Dong*, J. Wu, and G. Shi
    IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023 (TPAMI)
     [Paper] [Project Page] [code] [demo]

  7. Vector Quantization with Self-attention for Quality-independent Representation Learning
    We opt to learn invariant representations by learning image-quality-independent feature representation in a simple plug-and-play manner.
    Z. Yang, W. Dong*, X. Li, Y. Sun, M. Huang and G. Shi
    Conference on Computer Vision and Pattern Recognition, 2023 (CVPR)
     [Paper] [Project Page] [code] [demo]

  8. Self-supervised Non-uniform Kernel Estimation with Flow-based Motion Prior for Blind Image Deblurring
    We propose to represent the motion blur kernels in a latent space by a normalizing flow, and estimate the blur kernels in a self-supervised manner.
    Z. Fang, F. Wu*, W. Dong, X. Li, J. Wu, and G. Shi
    Conference on Computer Vision and Pattern Recognition, 2023 (CVPR)
     [Paper] [Project Page] [code] [demo]

  9. Adaptive Search-and-Training for Robust and Efficient Network Pruning
    We challenge the conventional wisdom of training before pruning by proposing a joint search-and-training approach to learn a compact network directly from scratch.
    X. Lu, W. Dong*, X. Li, J. Wu, L Li, and G. Shi
    IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023 (TPAMI)
     [Paper] [Project Page] [code] [demo]

  10. Deep unfolding network for efficient mixed video noise removal
    We introduce a deep image denoiser prior and obtain an iterative optimization algorithm based on the maximum a posterior (MAP) estimation.
    L. Sun, Y. Wang, F. Wu, X. Li, W. Dong, and G. Shi
    IEEE Transactions on Circuits and System for Video Technology, 2023 (T-CSVT)
     [Paper] [Project Page] [code] [demo]

  11. Searching Efficient Model-Guided Deep Network for Image Denoising
    We present a novel approach to fill this gap in image denoising application by connecting model-guided design (MoD) with NAS (MoD-NAS).
    Q. Ning, W. Dong*, X. Li, and J. Wu
    IEEE Transactions on Image Processing, 2023 (TIP)
     [Paper] [Project Page] [code] [demo]

  12. Differentiable Neural Architecture Search for Extremely Lightweight Image Super-Resolution
    We propose a novel differentiable Neural Architecture Search (NAS) approach on both the cell-level and network-level to search for lightweight SISR models.
    H. Huang, L. Shen, C. He, W. Dong and W. Liu
    IEEE Transactions on Circuits and System for Video Technology, 2023 (TCSVT)
     [Paper] [Project Page] [code] [demo]

  13. Supervised Contrastive Learning Based on Fusion of Global and Local Features for Remote Sensing Image Retrieval
    Supervised contrastive learning based on the fusion of global and local features method is proposed in this article, named SCFR.
    M. Huang, L. Dong, W. Dong and G. Shi
    IEEE Trans. Geosci. Remote. Sens, 2023 (TGRS)
     [Paper] [Project Page] [code] [demo]

  14. Bayesian deep learning for image reconstruction: from structured sparsity to uncertainty estimation
    We construct a new class of uncertainty-driven loss (UDL) functions for deep unfolded networks.
    W. Dong, J. Wu, L. Li, G. Shi, and X. Li
    IEEE Signal Processing Magazine, 2023 (SPM)
     [Paper] [Project Page] [code] [demo]

2022

  1. Uncertainty learning in kernel estimation for multi-stage blind image super-resolution
    A novel kernel estimation method was proposed with uncertainty learning, achieving SOTA blind image SR results.
    Z. Fang, W. Dong*, X. Li, J. Wu, L. Li, and G. Shi
    European Conference on Computer Vision, 2022 (ECCV)
     [Paper] [Project Page] [code] [demo]

  2. Self-feature distillation with uncertainty modeling for degraded image recognition
    A weighted feature distillation loss with uncertainty learning was proposed for degraded image recognition.
    Z. Yang, W. Dong*, X. Li, J. Wu, L. Li, and G. Shi
    European Conference on Computer Vision, 2022 (ECCV)
     [Paper] [Project Page] [code] [demo]

  3. Bayesian based re-parameterization for DNN model pruning
    We present a novel perspective of re-parametric pruning by Bayesian estimation.
    X. Lu, T. Xi, B. Li, G. Zhang, and W. Dong
    ACM Multimedia, 2022 (ACMM)
     [Paper] [Project Page] [code] [demo]

  4. Learning Degradation Uncertainty for Unsupervised Real-world Image Super-Resolution
    We propose a novel approach called USR-DU for unsupervised real-world image SR with learned degradation uncertainty.
    Q. Ning, J. Tang, F. Wu, W. Dong*
    International Joint Conferences on Artificial Intelligence, 2022 (IJCAI)
     [Paper] [Project Page] [code] [demo]

  5. Deep hyperspectral image fusion network with iterative spatio-spectral regularization
    We propose a novel regularization strategy to fully exploit the spatio-spectral dependency by a spatially adaptive 3D filter.
    T. Huang, W. Dong*, J. Wu, L. Li, X. Li, and Guangming Shi
    IEEE Transactions on Computational Imaging, 2022 (TCI)
     [Paper] [Project Page] [code] [demo]

  6. Robust depth completion with uncertainty-driven loss functions
    We introduce uncertainty-driven loss functions to improve the robustness of depth completion and handle the uncertainty in depth completion.
    Y. Zhu, W. Dong*, X Li, J. Wu, L. Li, and G. Shi
    Association for the Advancement of Artificial Intelligence, 2022 (AAAI)
     [Paper] [Project Page] [code] [demo]

2021

  1. Deep Maximum a Posterior Estimator for Video Denoising
    We present a novel deep maximum a posterior (MAP)-based video denoising method, with adaptive temporal fusion and deep image prior.
    L. Sun, W. Dong*, X. Li, J. Wu, L. Li, and G. Shi
    International Journal of Computer Vision, 2021 (IJCV)
     [Paper] [Project Page] [code] [demo]

  2. Model-Guided Deep Hyperspectral Image Super-resolution
    We propose to connect these two lines of research by presenting a Model-guided DCN (MoG-DCN) approach to HSISR.
    W. Dong, C. Zhou, F. Wu, J. Wu, G. Shi, and X. Li
    IEEE Transactions on Image Processing, 2021 (TIP)
     [Paper] [Project Page] [code] [demo]

  3. Deep Gaussian Scale Mixture Prior for Spectral Compressive Imaging
    We propose an interpretable HSI reconstruction method with learned Gaussian Scale Mixture (GSM) prior.
    T. Huang, W. Dong*, X. Yuan*, J. Wu, and G. Shi
    Conference on Computer Vision and Pattern Recognition, 2021 (CVPR)
     [Paper] [Project Page] [code] [demo]

  4. Uncertainty-driven loss for single image super-resolution
    We propose a new adaptive weighted loss for SISR to train deep networks focusing on challenging situations such as textured and edge pixels with high uncertainty.
    Q. Ning, W. Dong*, X. Li, J. Wu, and G. Shi
    Conference and Workshop on Neural Information Processing Systems,2021 (NeurIPS)
     [Paper] [Project Page] [code] [demo]

  5. Bayesian correlation filter learning with Gaussian scale mixture model for visual tracking
    We propose a principled Bayesian correlation filter learning method using Gaussian scale mixture (GSM) model.
    Y. Cao, G. Shi, T. Zhang, W. Dong*, J. Wu, X. Xie, and X. Li
    IEEE Transactions on Circuit and Systems for Video Technology,2021 (T-CSVT)
     [Paper] [Project Page] [code] [demo]

  6. Accurate and lightweight image super-resolution with model-guided deep unfolding network
    We present and advocate an explainable approach toward SISR named model-guided deep unfolding network (MoG-DUN).
    Q. Ning, W. Dong*, G. Shi, L. Li and X. Li
    IEEE Journal of Selected Topics on Signal Processing,2021 (J-STSP)
     [Paper] [Project Page] [code] [demo]

2020

  1. Beyond network pruning: a joint search-and-training approach
    We propose a coarse-to-fine tuning strategy to iteratively sample and update compact sub-network to approximate the target network.
    X. Lu, H. Huang, W. Dong*, G. Shi, and X. Li
    International Joint Conferences on Artificial Intelligence, 2020 (IJCAI)
     [Paper] [Project Page] [code] [demo]

  2. Accelerating convolutional neural network via structured Gaussian scale mixture models: a joint grouping and pruning approach
    We propose a hybrid network compression technique for exploiting the prior knowledge of network parameters by Gaussian scale mixture (GSM) models.
    T. Huang, W. Dong*, J. Liu, F. Wu, G. Shi, and X. Li
    IEEE Journal of Selected Topics on Signal Processing, 2020 (J-STSP)
     [Paper] [Project Page] [code] [demo]

  3. Spatial-temporal Gaussian scale mixture modeling for foreground estimation
    We proposed a novel spatial-temporal Gaussian scale mixture (STGSM) model for foreground estimation.
    Q. Ning, W. Dong*, F. Wu, J. Wu, J. Lin, and G. Shi
    Association for the Advancement of Artificial Intelligence, 2020 (AAAI)
     [Paper] [Project Page] [code] [demo]

2019

  1. Deep spatial-spectral representation learning for hyperspectral image denoising
    We present a novel, deep-learning framework for 3-D HSI denoising.
    W. Dong, H. Wang, F. Wu, G. Shi, and X. Li
    IEEE Transactions on Computational Imaging , 2019 (TCI)
     [Paper] [Project Page] [code] [demo]

  2. Denoising Prior Driven Deep Neural Network for Image Restoration
    We first propose a denoising-based IR algorithm, whose iterative steps can be computed efficiently. Then, the iterative process is unfolded into a deep neural network, which is composed of multiple denoisers modules interleaved with back-projection (BP) modules that ensure the observation consistencies.
    Weisheng Dong*, P. Wang, W. Yin, G. Shi, F. Wu, and X. Lu
    IEEE Transactions on Pattern Analysis and Machine Intelligence , 2019 (TPAMI)
     [Paper] [Project Page] [code] [demo]

2018 & Before

  1. Image Super-resolution with Parametric Sparse Model Learning
    We propose to take a hybrid approach toward image SR by combining those two lines of ideas-that is, a parametric sparse prior of HR images is learned from the training set as well as the input LR image.
    Y. Li, Weisheng Dong*, X. Xie, G. Shi, J. Wu, and X. Li
    IEEE Transactions on Image Processing , 2018 (TCI)
     [Paper] [Project Page] [code] [demo]

  2. Robust Foreground Estimation via Structured Gaussian Scale Mixture Modeling
    We propose to model the sparse component with a Gaussian scale mixture (GSM) model.
    G. Shi, T. Huang, Weisheng Dong*, J. Wu, and X. Xie
    IEEE Transactions on Image Processing , 2018 (TIP)
     [Paper] [Project Page] [code] [demo]

  3. Robust tensor approximation with Laplacian scale mixture modeling for multiframe image and video denoising
    We propose a novel robust tensor approximation (RTA) framework with the Laplacian Scale Mixture (LSM) modeling for three-dimensional (3-D) data and beyond
    Weisheng Dong, T. Huang, G. Shi, Y. Ma, and X. Li
    IEEE Journal of Selected Topics on Signal Processing , 2018 (J-STSP)
     [Paper] [Project Page] [code] [demo]

  4. Mixed noise removal via Laplacian scale mixture modeling and nonlocal low-rank approximation
    We propose an effective mixture noise removal method based on Laplacian scale mixture (LSM) modeling and nonlocal low-rank regularization.
    Tao Huang, Weisheng Dong*, Xuemei Xie, Guangming Shi, and Xiang Bai
    IEEE Transactions on Image Processing , 2017 (TIP)
     [Paper] [Project Page] [code] [demo]

  5. Color-guided depth recovery via joint local structural and nonlocal low-rank regularization
    We propose a unified variational approach via joint local and nonlocal regularization.
    Weisheng Dong, Guangming Shi, Xin Li, K. Peng, J. Wu, and Z. Guo
    IEEE Transactions on Multimedia , 2017 (TMM)
     [Paper] [Project Page] [code] [demo]

  6. Learning parametric sparse models for image super-resolution
    We propose to combine those two lines of ideas for image super-resolution.
    Y. Li, W. Dong*, X. Xie, G. Shi, X. Li, and D. Xu
    Conference and Workshop on Neural Information Processing Systems , 2016 (NeurIPS)
     [Paper] [Project Page] [code] [demo]

  7. Hyperspectral Image Super-Resolution via Non-Negative Structured Sparse Representation
    We propose a new hyperspectral image super-resolution method from a low-resolution (LR) image and a HR reference image of the same scene.
    Weisheng Dong, Fazuo Fu, Guangming Shi, and Xun Cao, Jinjian Wu, Guangyu Li, and Xin Li
    IEEE Transactions On Image Processing , 2016 (TIP)
     [Paper] [Project Page] [code] [demo]

  8. Image Restoration via Simultaneous Sparse Coding: Where Structured Sparsity Meets Gaussian Scale Mixture
    We propose a structured sparse coding framework.
    W. Dong, G. Shi, Y. Ma, and X. Li
    International Journal of Computer Vision , 2015 (IJCV)
     [Paper] [Project Page] [code] [demo]

  9. Low-rank tensor approximation with Laplacian scale mixture modeling for multiframe image denoising
    We propose a novel low-rank tensor approximation framework with Laplacian Scale Mixture (LSM) modeling for multi-frame image denoising.
    Weisheng Dong, Guangyu Li, Guangming Shi, Xin Li, and Yi Ma
    International Conference on Computer Vision , 2015 (ICCV)
     [Paper] [Project Page] [code] [demo]

  10. Learning parametric distributions for image super-resolution: where patch matching meets sparse coding
    We propose to develop a hybrid approach toward SR by combining those two lines of ideas.
    Yongbo Li, Weisheng Dong*, Guangming Shi, and Xuemei Xie
    International Conference on Computer Vision , 2015 (ICCV)
     [Paper] [Project Page] [code] [demo]

  11. Compressive sensing via nonlocal low-rank regularization
    We propose a nonlocal low-rank regularization (NLR) approach toward exploiting structured sparsity and explore its application into CS of both photographic and MRI images.
    W. Dong, G. Shi, X. Li, Y. Ma, and F. Huang
    IEEE Transactions on Image Processing , 2014 (TIP)
     [Paper] [Project Page] [code] [demo]