检索结果-内蒙古大学图书馆

Uni-COAL: A unified framework for cross-modality synthesis and super-resolution of MR images

EXPERT SYSTEMS WITH APPLICATIONS 2025年 270卷

作者： Song, Zhiyun Qi, Zengxin Wang, Xin Zhao, Xiangyu Shen, Zhenrong Wang, Sheng Fei, Manman Wang, Zhe Zang, Di Chen, Dongdong Yao, Linlin Liu, Mengjun Wang, Qian Wu, Xuehai Zhang, Lichi Shanghai Jiao Tong Univ Natl Engn Res Ctr Adv Magnet Resonance Technol Dia Sch Biomed Engn Shanghai 200030 Peoples R China Fudan Univ Huashan Hosp Shanghai Med Coll Dept Neurosurg Shanghai 200040 Peoples R China Natl Ctr Neurol Disorders Shanghai 200040 Peoples R China Shanghai Key Lab Brain Funct & Restorat & Neural R Shanghai 200040 Peoples R China Fudan Univ State Key Lab Med Neurobiol Shanghai 200040 Peoples R China Fudan Univ Frontiers Ctr Brain Sci Sch Basic Med Sci MOE Shanghai 200040 Peoples R China Fudan Univ Sch Basic Med Sci Shanghai 200040 Peoples R China Fudan Univ Inst Brain Sci Shanghai 200040 Peoples R China ShanghaiTech Univ Sch Biomed Engn Shanghai 201210 Peoples R China

Magnetic resonance imaging (MRI) enhancement techniques, such as cross-modality synthesis (CMS), super- resolution (SR), and their combination (CMSR), are crucial to improve the quality of images produced by simplified scanning protocols. However, existing approaches are typically tailored to specific tasks, limiting their flexibility and generalizability to handle complex clinical scenarios. Moreover, these MRI enhancement methods often treat alias frequencies improperly, resulting in suboptimal detail restoration performance. In this paper, we propose a Unified CO-modulated ALias-free framework ( Uni-COAL ) to accomplish the aforementioned tasks with a single network, so that the interaction of these tasks can be established and the resources for training and deploying the models can be greatly reduced. Specifically, the co-modulation design of the image- conditioned and stochastic attribute representations ensures the task consistency between CMS and SR, while simultaneously accommodating arbitrary combinations of modalities and thickness. The generator of Uni-COAL is also designed to be alias-free based on the Shannon-Nyquist signal processing framework, ensuring effective suppression of alias frequencies. Additionally, we leverage the semantic prior of the Segment Anything Model (SAM) to guide Uni-COAL, ensuring a more authentic preservation of anatomical structures during synthesis. Experiments on four datasets demonstrate that Uni-COAL outperforms the alternatives in various combinations of CMS, SR, and CMSR tasks, which highlights its superiority and flexibility to wide-range applications. Our codes are publicly available at https://***/zhiyuns/UniCOAL.

关键词： Cross-modality synthesis Super-resolution Magnetic resonance imaging

来源：评论

学校读者我要写书评

暂无评论

Research and design of image style transfer technology based on multi-scale convolutional neural network feature fusion

引用

ELECTRONICS LETTERS 2024年第11期60卷 e13250-e13250页

作者： Liang, Yongzhen Lin, Feng Xie, Wensheng Wang, Jinzhong Nie, Tian Guangxi Technol Coll Machinery & Elect Sch Artificial Intelligence Technol Nanning Peoples R China Guangxi Technol Coll Machinery & Elect Sch Green Bldg & Low Carbon Technol Nanning Peoples R China

To reduce the occurrence of information loss and distortion in image style transfer, a method is proposed for researching and designing image style transfer technology based on multi-scale convolutional neural networks (CNNs) feature fusion. Initially, the VGG19 model is designed for coarse and fine-scale networks to achieve multi-scale CNN feature extraction of target image information. Subsequently, while setting the corresponding feature loss function, an additional least-squares penalty parameter is introduced to balance the optimal total loss function. Finally, leveraging the characteristics of stochastic gradient descent iteration, image features are fused and reconstructed to obtain better style transfer images. Experimental evaluations utilize peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), information entropy (IE), and mean squared error (MSE) as metrics for assessing the transferred images, comparing them with three typical image style transfer methods. Results demonstrate that the proposed method achieves optimal performance across all metrics, realizing superior image style transfer effects. This article conducts theoretical research on the research and design of image style transfer technology based on multi-scale convolutional neural network (CNN) feature fusion, and compares and analyzes the implementation effects of various experimental methods. Based on the original CNN image style transfer method, improvements and designs are made, and the image style transfer effect is well achieved. image

关键词： image and vision processing and display technology image fusion image processing

来源：评论

学校读者我要写书评

暂无评论

Adversarial Attack with Adaptive Gradient Variance for Deep Fake Fingerprint Detection 24

Adversarial Attack with Adaptive Gradient Variance for Deep ...

引用

IEEE 24th International Workshop on Multimedia signal processing (MMSP)

作者： Yuan, Chengsheng Cui, Baojie Nanjing Univ Informat Sci & Technol Sch Comp Sci Minist Educ Engn Res Ctr Digital Forens Nanjing 210044 Peoples R China

ISBN: (数字)9781665471893

ISBN: (纸本)9781665471893

In recent years, fingerprint liveness detection methods based on deep neural networks have been widely used in various fingerprint recognition systems, such as fingerprint locks and smart phones, etc. However, recent research has revealed that these devices are vulnerable to the spoofing attack of adversarial examples. That is, adding some invisible perturbations to fingerprint image, the deep fake fingerprint detection network may misclassify the image as other classes. The classical example generation methods usually adopt gradient optimization method, but it is prone to falling into local optimum, resulting in poor adversarial attack performance. Accordingly, this paper proposes a novel adversarial attack method called adaptive gradient variance attack (AGVA) for deep fake fingerprint detection. Firstly, we obtain gradient information through the target network to calculate the gradient variance. Then, to avoid falling into the local optimum, the gradient direction is optimized by the gradient variance. In particular, to improve the attack performance, we propose an adaptive hyper-parameter search algorithm using stochastic gradient ascent to search the value of hyper-parameters in adversarial example generation. Experimental results show that compared to existing methods, our method achieves a better black-box attack success rate and higher robustness on the basis of ensuring the white-box attack performance.

关键词： Deep neural network fake fingerprint detection adversarial examples adaptive gradient variance

来源：评论

学校读者我要写书评

暂无评论

RETHINKING UNSUPERVISED neural SUPERPIXEL SEGMENTATION 29

RETHINKING UNSUPERVISED NEURAL SUPERPIXEL SEGMENTATION

引用

IEEE International Conference on image processing (ICIP)

作者： Eliasof, Moshe Ben Zikri, Nir Treister, Eran Ben Gurion Univ Negev Dept Comp Sci Beer Sheva Israel

ISBN: (数字)9781665496209

ISBN: (纸本)9781665496209

Recently, the concept of unsupervised learning for superpixel segmentation via CNNs has been studied. Essentially, such methods generate superpixels by convolutional neural network (CNN) employed on a single image, and such CNNs are trained without any labels or further information. Thus, such approach relies on the incorporation of priors, typically by designing an objective function that guides the solution towards a meaningful superpixel segmentation. In this paper we propose three key elements to improve the efficacy of such networks: (i) the similarity of the soft superpixelated image compared to the input image, (ii) the enhancement and consideration of object edges and boundaries and (iii) a modified architecture based on atrous convolution, which allow for a wider field of view, functioning as a multi-scale component in our network. By experimenting with the BSDS500 dataset, we find evidence to the significance of our proposal, both qualitatively and quantitatively. (1)

关键词： Superpixels Deep Learning

来源：评论

学校读者我要写书评

暂无评论

Multiband remote 3D-ViT data fusion 30

Multiband remote 3D-ViT data fusion

引用

Conference on Algorithms, Technologies, and Applications for Multispectral and Hyperspectral Imaging XXX

作者： Kurz, Aidan G. Depoian, Arthur C., ii Bailey, Colleen P. Univ North Texas Dept Elect Engn Denton TX 76207 USA

ISBN: (纸本)9781510673816;9781510673809

Satellites are equipped with diverse sensors, capable of capturing detailed information across a multitude of wavelengths. The fusion of multispectral data is pivotal to amplify the visual representation of the area of interest. The improvement of information representation allows for enhanced processing, analysis, and other crucial tasks for numerous fields of study, including remote sensing, defense, and material characterization. Previous solutions often utilize traditional signal processing techniques, including principal component analysis (PCA), to accomplish data fusion. By performing fusion on a feature level, extracted information about the area of interest texture and boundaries are combined. The introduction of neural network techniques improved the reconstruction of data similar to the results obtained by conventional inference of humans. For example, the use of deep learning algorithms in conjunction with PCA allowed for refined reduction of redundancy and distortion of spectral data, in comparison to traditional methods alone. The introduction of the Vision Transformer (ViT) architecture, originally developed for two-dimensional image data, has revolutionized image processing tasks, vastly improving performance at the cost of a large quantity of trainable parameters. Recent experimentation has proven that optimizing ViT for efficiency allows for comparable or even superior performance while lessening the computational cost. The transition from 2D to 3D information via utilization of additional depth and spatial data has also led to superior results as the added information allows for better representation of terrain features, making it invaluable for satellite imagery analysis. Combining the principles of ViT and 3D information to process complex satellite data can result in more effective data fusion to achieve a superior level of data visualization of multispectral satellite imagery in an efficient manner.

关键词： Data Fusion Multispectral Fusion Remote Sensing Deep Learning

来源：评论

学校读者我要写书评

暂无评论

A Review of image Enhancement methods Based on stochastic Resonance of Neuronal Networks

A Review of Image Enhancement Methods Based on Stochastic Re...

引用

Pattern Recognition and Machine Learning (PRML), IEEE International Conference on

作者： Ximin Sun Pingjuan Niu Yong Li Xiaoming Li Bingzhen Zhao Bin Zhang Dan Liu Bo Sun Tiangong University Laboratory of Metaverse in Electric Power Industry State Gride Ecommerce Technology CO.LTD Tian'jin China Laboratory of Metaverse in Electric Power Industry Tiangong University Tian'jin China Laboratory of Metaverse in Electric Power Industry State Gride Ecommerce Technology CO.LTD Tian'jin China

ISBN: (数字)9798350355925

ISBN: (纸本)9798350355932

stochastic resonance has always been a research hotspot in the field of image processing. The ability of the nervous system to detect signals is closely related to nonlinear and collective behavior in neural mediators. Neuronal noise is the most significant factor influencing this ability, and it is a natural endogenous fluctuation in brain activity. Experimental and theoretical studies have shown that noise is beneficial, not harmful, for the normal functioning of the nervous system. In this regard, it is generally agreed that noise of appropriate intensity can participate in the rhythmic activity of the brain, and that noise-induced oscillations enhance the performance of weak signal processing, especially when the frequency of the signal is approximately noise-induced rhythmic oscillations. This behavior in the biological nervous system can be explained by the concept of “stochastic resonance”. Correspondingly, the stochastic resonance mechanism of neurons is used to denoise the image, which has a good effect. In this paper, we first introduce an image processing method using random resonance of neurons. Then, the commonly used evaluation indicators in the field of image processing are sorted out. Finally, the development trend of image processing using random resonance of neurons is prospected.

关键词： Reviews stochastic resonance Noise Neurons Resonant frequency Machine learning Market research Pattern recognition Oscillators image enhancement

来源：评论

学校读者我要写书评

暂无评论

On the Role of Adversarial Perturbations in Unsupervised Deep Learning for Enhancing image Quality of Degraded Medical-images 22

On the Role of Adversarial Perturbations in Unsupervised Dee...

引用

22nd IEEE International Symposium on Biomedical Imaging, ISBI 2025

作者： Sharma, Vatsala Goyal, Ashwin Awate, Suyash P. Indian Institute of Technology Bombay Computer Science and Engineering Department Mumbai India

ISBN: (纸本)9798331520526

Unsupervised learning enables deep-neural-network (DNN) training without high-quality data. Typical DNN methods for unsupervised image quality enhancement either (i) focus only on spatially uncorrelated noise, (ii) demand a training set comprising multiple noisy acquisitions per scene, or (iii) apply only to specific data;these aspects limit their applicability to medical images in clinical applications. Moreover, most such existing methods lack robustness to out-of-distribution (OOD) images observed during deployment. We propose a novel unsupervised DNN learning formulation to enhance a variety of medical images. Our unsupervised DNN training optimizes sets of multistage adversarial perturbations to (i) auto-generate multiple degraded inputs and latent-space encodings for each training image, (ii) adapt to the textural statistics of real-world degradations across medical imaging domains, and (iii) gain robustness to OOD images. This formulation also includes uncertainty modeling and estimation. Results on three publicly available datasets (CT, PET, MRI) show the benefits of our method in image-quality enhancement on in-distribution and OOD images. © 2025 IEEE.

关键词： Adversarial perturbations Medical images Out-of-distribution Uncertainty Unsupervised learning

来源：评论

学校读者我要写书评

暂无评论

Nonlinear Transform Coding

引用

IEEE JOURNAL OF SELECTED TOPICS IN signal processing 2021年第2期15卷 339-353页

作者： Balle, Johannes Chou, Philip A. Minnen, David Singh, Saurabh Johnston, Nick Agustsson, Eirikur Hwang, Sung Jin Toderici, George Google Res Mountain View CA 94043 USA

We review a class of methods that can be collected under the name nonlinear transform coding (NTC), which over the past few years have become competitive with the best linear transform codecs for images, and have superseded them in terms of rate-distortion performance under established perceptual quality metrics such as MS-SSIM. We assess the empirical rate-distortion performance of NTC with the help of simple example sources, for which the optimal performance of a vector quantizer is easier to estimate than with natural data sources. To this end, we introduce a novel variant of entropy-constrained vector quantization. We provide an analysis of various forms of stochastic optimization techniques for NTC models;review architectures of transforms based on artificial neural networks, as well as learned entropy models;and provide a direct comparison of a number of methods to parameterize the rate-distortion trade-off of nonlinear transforms, introducing a simplified one.

关键词： Transforms Entropy Distortion image coding Vector quantization stochastic processes Artificial neural networks data compression machine learning rate-distortion source coding transform coding unsupervised learning

来源：评论

学校读者我要写书评

暂无评论

An Innovative Approach for Long ECG Synthesis with Wasserstein GAN Model 24th

An Innovative Approach for Long ECG Synthesis with Wasserste...

引用

24th International Conference on Computational Science and Its Applications (ICCSA)

作者： Tran, Thi Diem Dang, Thi Thu Khiet Tran, Ngoc Quoc Univ Informat Technol VNU HCM Ho Chi Minh City 70000 Vietnam Univ Int VNU HCM Ho Chi Minh City 70000 Vietnam Univ Sci John Von Neumann Inst VNU HCM Ho Chi Minh City 70000 Vietnam

ISBN: (纸本)9783031646072;9783031646089

Deep neural networks (DNNs) have set new standards in identifying and classifying irregular patterns in ECG (electrocardiogram) signals, surpassing previous methods. Despite the easy access and affordability of ECG sensors, a critical bottleneck remains the limited availability of reliable data for complex heart rhythms like second and third-degree atrioventricular block, ventricular tachycardia, and supraventricular tachycardia. This shortage has been a significant obstacle to improving DNN algorithms. Recent studies have turned to Generative Adversarial Networks (GANs) to create synthetic ECG data, enhancing the diversity of training datasets. However, much of this research has only managed to produce basic ECG components, missing the intricate details found in real patient data that includes multiple heartbeats. Our research has taken a groundbreaking approach by converting ECG signals into a two-dimensional format, allowing us to utilize advanced GAN models originally developed for image processing. This method has enabled us to generate extended, realistic ECG sequences closely mimicking those from actual patients. We have tested and refined our model using two databases, Physionet and Chapman, and have successfully produced 10-second ECG sequences showcasing a variety of heart rhythms previously unachieved in other studies. Our innovative technique not only surpasses existing methods in generating high-quality, realistic ECG data but also sets a new benchmark in ECG synthesis.

关键词： ECG generating Long ECG signal to image GAN

来源：评论

学校读者我要写书评

暂无评论

MULTI-SCALE RAFT: COMBINING HIERARCHICAL CONCEPTS FOR LEARNING-BASED OPTICAL FLOW ESTIMATION 29

MULTI-SCALE RAFT: COMBINING HIERARCHICAL CONCEPTS FOR LEARNI...

引用

IEEE International Conference on image processing (ICIP)

作者： Jahedi, Azin Mehl, Lukas Rivinius, Marc Bruhn, Andres Univ Stuttgart Inst Visualizat & Interact Syst Stuttgart Germany Univ Stuttgart Inst Informat Secur Stuttgart Germany

ISBN: (数字)9781665496209

ISBN: (纸本)9781665496209

Many classical and learning-based optical flow methods rely on hierarchical concepts to improve both accuracy and robustness. However, one of the currently most successful approaches - RAFT - hardly exploits such concepts. In this work, we show that multi-scale ideas are still valuable. More precisely, using RAFT as a baseline, we propose a novel multi-scale neural network that combines several hierarchical concepts within a single estimation framework. These concepts include (i) a partially shared coarse-to-fine architecture, (ii) multi-scale features, (iii) a hierarchical cost volume and (iv) a multi-scale multi-iteration loss. Experiments on MPI Sintel and KITTI clearly demonstrate the benefits of our approach. They show not only substantial improvements compared to RAFT, but also state-of-the-art results - in particular in non-occluded regions. Code will be available at https://***/cv-stuttgart/MS_RAFT.

关键词： Optical flow neural networks Coarse-to-fine estimation Multi-scale features Multi-scale loss

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：