检索结果-内蒙古大学图书馆

Exploring fusion domain: Advancing infrared and visible image fusion via IDFFN-GAN

NEUROCOMPUTING 2025年 611卷

作者： Li, Juan Shi, Xiaoqian Li, Yanan Zhou, Huabing Wuhan Inst Technol Wuhan 430250 Peoples R China Wuhan Inst Technol Hubei Key Lab Intelligent Robot Wuhan 430250 Peoples R China

Infrared (IR) and visible (VI) images are crucial in applications such as surveillance and night vision, where each modality provides complementary information-IR captures thermal details, while VI captures textures. Fusing these images is essential to combine their strengths, resulting in a more comprehensive and informative image. In this work, we introduce the "fusion domain" concept, a unique distribution that optimally blends IR and VI features. Our primary contribution is developing the Intermediate Domain Feature Fusion Network (IDFFN), which employs an MLP to learn the optimal domain factor for this fusion. Integrated with the IDFFN-GAN's dual discriminators, our approach refines the fusion process to produce images that better preserve thermal and texture details from the proposed fusion and gradient domains. Experimental results demonstrate that our method achieves a 5% boost in Entropy (EN), a 50% rise in Spatial Frequency (SF), a 58% improvement in Average Gradient (AG), a 10% enhancement in Standard Deviation (SD), and a remarkable 198% increase in N AB / F compared to state-of-the-art techniques, ensuring superior retention of specific features from both modalities.

关键词： Image fusion Generative adversarial network Intermediate domain Domain factor Dual discriminators

来源：评论

学校读者我要写书评

暂无评论

Single Image Dehazing via Multi-Scale Large Kernel Convolutional Neural Networks

Single Image Dehazing via Multi-Scale Large Kernel Convoluti...

引用

International Joint Conference on Neural Networks (IJCNN)

作者： Li, Minghui Liu, Wei Kang, Zhiguo Huang, Xiaoyu Wuhan Inst Technol Hubei Key Lab Intelligent Robot Wuhan Peoples R China

ISBN: (纸本)9798350359329;9798350359312

Large-kernel convolutional neural networks (CNNs) have recently achieved remarkable performance comparable to Visual Transformers(ViTs) in high-level vision tasks. However, there are two critical drawbacks hindering its widespread applications in image dehazing. 1) Most large-kernel designs focus expanding the kernel size even further to model stronger long-range dependencies, but this approach brings a substantial amount of computational overhead. 2) As the kernel size increases, the network tends to focus more on the shape of the object over its texture, potentially affecting the details of the recovered image. To overcome these issues, we propose an effective multi-scale large separable kernel attention module (MLSKA) that can simultaneously build long-range and local dependencies in a cost-effective manner to facilitate high-quality image reconstruction. Specifically, MLSKA combines an efficient convolutional decomposition design with multi-scale learning, realizing multi-scale receptive fields while significantly reducing the computational cost and parameters of large-kernel convolution. In addition, we introduce a deformable attention feedforward network (DAFN) to aggregate contextual information. In DAFN, a novel deformable attention gate is designed to provide holistic attention to the feed-forward network (FFN), thereby improving its utilization of critical features. Integrating these two designs into a U-shaped backbone, the proposed multi-scale large-kernel network (MLANet) outperforms state-of-the-art methods on several dehazing benchmarks, achieving the best parameter-performance trade-off.

关键词： Image Dehazing Convolutional Neural Networks Large Kernel Attention Mechanism

来源：评论

学校读者我要写书评

暂无评论

Rain Streak Attention Network for Single Image Deraining

Rain Streak Attention Network for Single Image Deraining

引用

International Joint Conference on Neural Networks (IJCNN)

作者： Kang, Zhiguo Liu, Wei Zhang, Caiwang Li, Minghui Wuhan Inst Technol Hubei Key Lab Intelligent Robot Wuhan Peoples R China

ISBN: (纸本)9798350359329;9798350359312

Single image deraining methods have been extensively studied for its ability to remarkably improve the performance of computer vision tasks in rainy environments. However, existing methods are rarely able to use the properties of the rain streaks themselves to remove rain, and they often have difficulty correctly identifying the rain streaks and background information when removing complex rain streaks, resulting in the loss of some of the detailed textures. In this paper, we propose a Rain Streak Attention Network, which can make full use of rain features between different color channels to identify rain streaks from complex backgrounds. The basic building blocks of the method are the rain streak extraction block (RSEB) and the detailed feature recovery block (DFRB). Specifically, RSEB aims to separate rainbands from background information by focusing on similar rainfall characteristics in different channels using an attention mechanism. In DFRB, we achieve the modeling of global features by deep-wise dilation convolution, which helps to recover more realistic detailed texture information. Extensive experimental results on the commonly used benchmarks demonstrate that the proposed method achieves favorable performance against state-of-the-art approaches.

关键词： Image Deraining Attention

来源：评论

学校读者我要写书评

暂无评论

Deep single image deraining using an asymmetric cyclic generative and adversarial framework

引用

APPLIED INTELLIGENCE 2024年第8期54卷 6776-6798页

作者： Liu, Wei Zhang, Caiwang Chen, Cheng Huang, Xiaoyu Li, Minghui Wuhan Inst Technol Sch Comp Sci & Engn Wuhan 430205 Hubei Peoples R China Wuhan Inst Technol Hubei Key Lab Intelligent Robot Wuhan 430205 Hubei Peoples R China

In reality, rain and fog are often present simultaneously, which can greatly reduce the clarity and quality of scene images. However, most unsupervised single image deraining methods primarily concentrate on removing rain streaks and disregard the presence of fog, which often leads to low deraining performance. In addition, these methods generate samples that are too similar and lack diversity, resulting in poor performance when dealing with complex rain scenes. To address the above issues, we propose a novel Asymmetric Cyclic Generative and Adversarial Framework (ACGF) for single image deraining in which the deraining model, which consists of a Rain-fog2Clean (R2C) transformation block and a Clean2Rain-fog (C2R) transformation block, is trained on both synthetic and real rainy images to simultaneously capture both rain streaks and fog features. To better characterize combined rain-fog features in the R2C block, we propose an attention-based rain-fog feature extraction (ARFE) network to exploit the self-similarity of global and local rain-fog information by learning spatial feature correlations. Furthermore, to improve the translational capacity of the C2R block and the diversity of the model on the synthetic rain conversion path, we design a rain-fog feature decoupling and reorganization (RFDR) network by embedding a rainy image degradation model and a mixed discriminator to preserve richer texture details. Extensive experiments on benchmark rain-fog, rain and fog datasets show that our ACGF outperforms state-of-the-art deraining methods.

关键词： Attention mechanism Unsupervised image deraining Adversarial learning

来源：评论

学校读者我要写书评

暂无评论

Cross-view Aggregation Network For Stereo Image Super-Resolution

Cross-view Aggregation Network For Stereo Image Super-Resolu...

引用

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

作者： Chen, Zhitao Lu, Tao Zhao, Kanghui Zhu, Bolin Li, Zhen Wang, Jiaming Zhang, Yanduo Wuhan Inst Technol Hubei Key Lab Intelligent Robot Wuhan Peoples R China

ISBN: (纸本)9798350365474

Although stereo image super-resolution has been extensively studied, many existing works only rely on attention in a single epipolar direction to reconstruct stereo images. In the case of asymmetric parallax images, these methods often struggle to capture reliable stereo correspondence, resulting in reconstructed images suffering from blurring and artifacts. In this paper, we propose a novel method called Cross-View Aggregation Network for Stereo Image Super-Resolution (CANSSR) and explore the relationship between multi-directional epipolar lines to construct reliable stereo correspondence. Specifically, we propose a multidirectional cross-view aggregation module (MCAM) that effectively captures multi-directional stereo correspondence and obtains cross-view complementary information. Furthermore, we design a channel-spatial aggregation module (CSAM) that aggregates multi-order global-local information in intra-view to reconstruct clearer texture features. In addition, we equip a large kernel convolution in the Feed-forward Network to acquire richer detailed texture information. The extensive experiments conclusively demonstrate that CANSSR outperforms the state-of-the-art method both qualitatively and quantitatively in terms of stereo image super-resolution on the Flickr 1024 and Middlebury datasets.

关键词： Stereo image processing

来源：评论

学校读者我要写书评

暂无评论

Pre-trained Bidirectional Dynamic Memory Network For Long Video Question Answering

Pre-trained Bidirectional Dynamic Memory Network For Long Vi...

引用

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

作者： Wu, Jinmeng Shu, Pengcheng Hong, Hanyu Ma, Lei Zhu, Ying Wang, Lei Wuhan Inst Technol Hubei Key Lab Intelligent Robot Wuhan Peoples R China

ISBN: (纸本)9798350365474

Video question and answering is an important problem that receives extensive research interests in past few years. Although visual language models have seen recent success in several multimodal tasks, they still face challenges when it comes to complex reasoning in long movies that involve multiple person-object interaction events. Humans tackle these issues by utilizing a sequence of episodic memories as reference points to swiftly identify crucial times that are pertinent to the task at hand. In order to emulate this successful method of reasoning, we propose a pre-trained bidirectional dynamic memory model for long video question answering. In the feature extraction stage, we use a pre-trained CLIP model to generate video features and question features. Then, we use a bidirectional matching memory network to acquire events memories. In the features fusion stage, the events memories will be used as connecting points to establish associations between the high-level event concepts and the low-level redundant video content using the spatio-temporal self-attention. When dealing with the problem, our model first prioritizes the generated key event memories and then considers the most relevant event memories. Our contribution addresses the difficulty of effectively matching video features and question features in complex video contexts in long video question answering, and where features are easily missing under long temporal video sequences. Our experiments on two public video question answering datasets, which proves that our model has good performance in long video question answering.

关键词： Long video question answering Memory network Multi-event reasoning

来源：评论

学校读者我要写书评

暂无评论

Hybrid Steganography: Leveraging Chaotic Encryption and CNN for Robust Image Hiding 7

Hybrid Steganography: Leveraging Chaotic Encryption and CNN ...

引用

7th International Conference on Mechatronics, robotics and Automation

作者： Niu, Xiaoqing Yi, Guohong Wuhan Inst Technol Hubei Key Lab Intelligent Robot Wuhan Peoples R China

ISBN: (纸本)9798350352481;9798350352474

This paper presents a robust image information hiding method integrating Convolutional Neural Networks (CNN) and image encryption based on the Lorenz three-dimensional chaotic system. We introduce chaotic image encryption technology based on the Lorenz three-dimensional system into the CNN-based image steganography framework. Leveraging the high sensitivity and unpredictability of the Lorenz system, the original secret image is first encrypted to render it visually imperceptible. During the image hiding phase, the encrypted secret image is embedded into the carrier image using an improved CNN-based image information hiding model. This model capitalizes on CNN's powerful automatic learning and feature extraction capabilities to achieve information hiding, resulting in a steganographic image. In the information extraction phase, the steganographic image serves as the input for the decoder network within the model, where the decoder decrypts it to retrieve both the carrier image and the encrypted secret image. Finally, the secret image is decrypted using the Lorenz chaotic encryption algorithm. Analyses of PSNR/SSIM and visual quality demonstrate that the proposed hybrid image steganography scheme offers high security and robustness, effectively preventing distortion of the secret image.

关键词： Image Hiding CNN Lorenz Chaotic System Image Encryption

来源：评论

学校读者我要写书评

暂无评论

Research on Dynamic Load Balancing Based on Artificial Bee Colony Algorithm 7

Research on Dynamic Load Balancing Based on Artificial Bee C...

引用

7th International Conference on Mechatronics, robotics and Automation

作者： Yang, Yi Yi, Guohong Wuhan Inst Technol Hubei Key Lab Intelligent Robot Wuhan Peoples R China

ISBN: (纸本)9798350352481;9798350352474

With the rapid development of the Internet, explosive access requests pose significant challenges to server performance and capacity. Considering the inefficiencies of current load balancing algorithms, such as decreased system efficiency at lower loads and suboptimal allocation efficiency at higher loads, this paper proposes a dynamic load balancing algorithm based on the Nginx server and the artificial bee colony algorithm. Through the artificial bee colony algorithm, the threshold for the conversion between static and dynamic load balancing is calculated. If the server load is below this threshold, a static algorithm is employed;otherwise, a dynamic algorithm is used. This algorithm utilizes server performance parameters such as CPU performance, memory performance, disk I/O, and network bandwidth as evaluation metrics. Through comparative experiments, the algorithm's effectiveness is demonstrated in experimental environments, showing superior load balancing effects compared to weighted round-robin and least connection algorithms, effectively improving average server response time and actual concurrent connection numbers.

关键词： Server Performance Threshold Load Balancing Artificial Bee Colony Algorithm

来源：评论

学校读者我要写书评

暂无评论

Reconstructing visible-near-infrared hyperspectral images via liquid crystal based optoelectronic hybrid neural network

引用

OPTICS AND LASER TECHNOLOGY 2025年 184卷

作者： Li, Shiqi Li, Hui Li, Tian Song, Xiaoyue Wu, Yuntao Wuhan Inst Technol Sch Comp Sci & Engn Wuhan 430205 Peoples R China Hubei Key Lab Intelligent Robot Wuhan 430205 Peoples R China

We introduce an Optoelectronic Hybrid Neural Network (OHNN) that integrates software and hardware optimizations to enable the simultaneous reconstruction of visible-near-infrared (VIS-NIR) hyperspectral images. The network employs a linear polarizer (LP) and liquid crystal (LC) microlens array (MLA). We harness the polarization and electrically controlled tunable properties of LC-MLA in conjunction with digital optics technology to construct a software-defined dispersion element digital prism (DP). The OHNN can reconstruct VIS-NIR hyperspectral images, ranging from 400 nm to 900 nm, from limited datasets with a spectral accuracy of 10 nm. The reconstructed hyperspectral images boast a Peak signal-to-noise ratio of approximately 31 dB, marking a 3.2 % improvement over existing state-of-the-art methods.

关键词： Visible-near-infrared hyperspectral reconstruction Optoelectronic Hybrid Neural Network Digital prism Liquid crystal microlens array

来源：评论

学校读者我要写书评

暂无评论

TripleA: An Unsupervised Domain Adaptation Framework for Nighttime VRU Detection

引用

IEEE TRANSACTIONS ON intelligent TRANSPORTATION SYSTEMS 2025年第6期26卷 8320-8336页

作者： Wang, Yuankun Shao, Zhenfeng Wang, Jiaming Wang, Yu Ding, Yulin Cheng, Gui Wuhan Univ State Key Lab Informat Engn Surveying Mapping & Re Wuhan 430079 Peoples R China Wuhan Inst Technol Hubei Key Lab Intelligent Robot Wuhan 430200 Peoples R China

Detecting vulnerable road users (VRUs) at night presents significant challenges. Numerous methods rely heavily on annotations, yet the low visibility of nighttime images poses difficulties for labeling. To obviate the need for nighttime annotations, unsupervised domain adaptation manifests as a viable solution. However, existing approaches primarily focus on semantic-level domain gaps, often overlooking pixel-level discrepancies caused by inherent degradations in the nighttime domain. These degradations can impair machine vision and limit detection performance. In this paper, we propose TripleA, an unsupervised domain adaptation framework tailored for nighttime VRU detection. TripleA includes triple alignment. First, it aligns daytime and nighttime images to generate synthetic nighttime images, which are then enhanced for illumination and noise. To remove noise, we introduce an illumination difference-aware denoising network, incorporating a novel pseudo-supervised attention to achieve pixel-wise noise distribution alignment. This alignment is driven by pseudo-ground truth generated through a carefully designed exchange-recombination strategy, facilitating self-supervised training of the denoising network. Additionally, we introduce degradation alignment to ensure domain-invariant degradation encoding, which enhances the network's robustness for real-world nighttime images. Extensive experiments demonstrate the effectiveness of our framework for nighttime VRU detection, all without the need for annotated nighttime data.

关键词： Lighting Noise Degradation Noise reduction Training Image enhancement Annotations Detectors Atmospheric modeling Roads Vulnerable road user detection unsupervised domain adaptation image enhancement low-light denoising

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：