检索结果-内蒙古大学图书馆

A total variation global optimization framework and its application on infrared and visible image fusion

signal image AND VIDEO processing 2022年第1期16卷 219-227页

作者： Gao, Zhisheng Wang, Qiaolu Zuo, Chenglin Xihua Univ Sch Comp & Software Engn Chengdu 610039 Peoples R China China Aerodynam Res & Dev Ctr Low Speed Aerodynam Inst Mianyang 621000 Sichuan Peoples R China

The main bottleneck faced by total variation methods for image fusion is that it is difficult to design a novel optimization model that can be solved by numerical methods. This paper proposes a general framework of total variation optimized by deep learning for infrared and visible image fusion, which combines the advantages of deep convolutional neural networks. Under this framework, any arbitrary convex or non-convex total variation model for image fusion can be designed, and its optimization solution can be obtained through neural network learning. The core idea of the proposed framework is to transform the designed variational model into a loss function of a deep convolutional neural network, and then use the initial fused image of a source image and the output fused image to represent the data item, use the output image and the source image to represent the regularization term, and finally use a deep neural network learning method to obtain the optimal fused image. Based on the proposed framework, further research on pre-fusion, network model and regularization item can be carried out. To verify the effectiveness of the proposed framework, we designed a specific non-convex total variational model and performed experiments on the infrared and visible image datasets. Experimental results show that the proposed method has strong robustness, and compared with the fused images obtained by current state-of-art algorithms in terms of objective evaluation metrics and visual effects, the fused image obtained by the proposed method has more competitive advantages. Our code is publicly available at https://***/gzsds/globaloptimizationimagefusion.

关键词： Deep learning Total variation image fusion Deep convolutional neural network Infrared and visible image fusion

来源：评论

学校读者我要写书评

暂无评论

BI-DIRECTIONAL TRACKLET EMBEDDING FOR MULTI-OBJECT TRACKING 31

BI-DIRECTIONAL TRACKLET EMBEDDING FOR MULTI-OBJECT TRACKING

引用

2024 International Conference on image processing

作者： Bilgi, H. Cagri Alatan, A. Aydin METU Dept Elect & Elect Engn Ankara Turkiye METU Ctr Image Anal OGAM Ankara Turkiye

ISBN: (纸本)9798350349405;9798350349399

The last decade has seen significant advancements in multi-object tracking, particularly with the emergence of deep learning based methods. However, many prior studies in online tracking have primarily focused on enhancing track management or extracting visual features, often leading to hybrid approaches with limited effectiveness, especially in scenarios with severe occlusions. Conversely, in offline tracking, there has been a lack of emphasis on robust motion cues. In response, this approach aims to present a novel solution for offline tracking by merging tracklets using some recent promising learning-based architectures. We leverage a jointly performing Transformer and Graph neural Network (GNN) encoder to integrate both the individual motions of targets and their interactions in between. By enabling bi-directional information propagation between the Transformer and the GNN, proposed model allows motion modeling to depend on interactions, and conversely, interaction modeling to depend on the motion of each target. The proposed solution is an end-to-end trainable model that eliminates the requirement for any handcrafted short-term or long-term matching processes. This approach performs on par with state-of-the-art multi-object tracking algorithms, demonstrating its effectiveness and robustness.

关键词： multi-object tracking tracklet embedding transformer graph neural network hierarchical tracking

来源：评论

学校读者我要写书评

暂无评论

Grassmannian learning mutual subspace method for image set recognition

引用

NEUROCOMPUTING 2023年 517卷 20-33页

作者： Souza, Lincon S. Sogi, Naoya Gatto, Bernardo B. Kobayashi, Takumi Fukui, Kazuhiro Natl Inst Adv Ind Sci & Technol Dept Informat Technol & Human Factors Koto Ku 2-4-7 Aomi Tokyo 1350064 Japan Univ Tsukuba Fac Engn Informat & Syst 1-1-1 Tennodai Tsukuba Ibaraki 3058573 Japan Univ Tsukuba Grad Sch Syst & Informat Engn 1-1-1 Tennodai Tsukuba Ibaraki 3058573 Japan

This paper addresses the problem of object recognition given a set of images as input (e.g., multiple cam-era sources and video frames). Convolutional neural network (CNN)-based frameworks do not exploit these sets effectively, processing a pattern as observed, not capturing the underlying feature distribution as it does not consider the variance of images in the set. To address this issue, we propose the Grassmannian learning mutual subspace method (G-LMSM), a NN layer embedded on top of CNNs that can process image sets more effectively and can be trained in an end-to-end manner. The image set is first represented by a low-dimensional input subspace and then this input subspace is matched with dic-tionary subspaces by a similarity of their canonical angles, an interpretable and easy to compute metric. The key idea of G-LMSM is that the dictionary subspaces are learned as points on the Grassmann man-ifold, optimized with Riemannian stochastic gradient descent. This learning is stable, efficient and theo-retically well-grounded. We demonstrate the effectiveness of our proposed method on hand shape recognition, face identification, and facial emotion recognition.(c) 2022 Elsevier B.V. All rights reserved.

关键词： Grassmannian learning mutual subspace method Learning subspace methods Subspace learning image recognition Deep neural networks Manifold optimization

来源：评论

学校读者我要写书评

暂无评论

Simultaneous single image super-resolution and blind Gaussian denoising via slim ghost full-frequency residual blocks

引用

IET image processing 2024年第14期18卷 4799-4823页

作者： Farhangfar, Saghar Baradarani, Aryaz Asadpour, Mohammad Balafar, Mohammad Ali Maev, Roman Gr. Univ Tabriz Fac Elect & Comp Engn Tabriz Iran Tessonics Inc Ctr Diagnost Imaging & Signal Proc Windsor ON Canada Univ Windsor Inst Diagnost Imaging Res Windsor ON Canada

Given that super-resolution (SR) aims to recover lost information, and low-resolution (LR) images in real-world conditions might be corrupted with multiple degradations, considering basic bicubic down-sampling as the sole degradation significantly limits the performance of most existing SR models. This paper presents a model for simultaneous super-resolution and blind additive white Gaussian noise (AWGN) denoising with two components (netdeg and netSR) that is based on a generative adversarial network (GAN) to achieve detailed results. netdeg, featuring residual and innovative cost-effective ghost residual blocks with a frequency separation module for obtaining long-range information, blindly restores a clean version of the LR image. netSR leverages slim ghost full-frequency residual blocks to process low-frequency (LF) and high-frequency (HF) information via static large convolutions and pixel-wise highlighted input-adaptive dynamic convolutions, respectively. To address the susceptibility of dynamic layers to noise and preserve feature diversity while reducing model's costs, static and dynamic layer features are combined and highlighted. Diverse and non-redundant features are then processed using ghost-style blocks. The proposed model achieves comparable SR results in bicubic down-sampling scenarios, outperform existing SR methods in the complex task of concurrent SR and AWGN denoising, and demonstrate robustness in handling images corrupted with varying levels of AWGN.

关键词： convolutional neural nets feature extraction image representation image resolution

来源：评论

学校读者我要写书评

暂无评论

A CROSS DOMAIN GENERATIVE NETWORK FOR ACCELERATED MRI 31

A CROSS DOMAIN GENERATIVE NETWORK FOR ACCELERATED MRI

引用

2024 International Conference on image processing

作者： Ibrahim, Vazim Paul, Joseph Suresh Cochin Univ Sci & Technol Ernakulam 682022 Kerala India Indian Inst Informat Technol & Management Kerala Trivandrum 695581 Kerala India Digital Univ Kerala Sch Informat Trivandrum 695317 Kerala India

ISBN: (纸本)9798350349405;9798350349399

Compressed sensing (CS) is a method for accelerating MRI acquisition by acquiring less data through undersampling of Fourier space (k-space). Existing deep learning-based CSMRI methods are commonly concerned with optimizing datadriven network models with input undersampled data points and an efficient learning framework. Generative modelling is a learning framework employed in different applications for learning an abstract distribution of observed data and thereby generating new data points similar to the true features. In this regard, the current work proposes a Generative Adversarial Network (GAN) based Cross Domain Extrapolation Generative Adversarial Network (CdE-GAN) that incorporates an extrapolation mechanism through decoder-type architectural design to represent the fine details with a large set of pixels. The results obtained from the experiments show that the extrapolation network enables robust and accurate estimation of missing frequencies, alleviating the structural artifacts at higher acceleration/downsampling factors compared to state-of-the-art methods.

关键词： Compressed Sensing MRI Super-Resolution MRI Generative Adversarial Network Cross-domain Networks Convolutional neural Network

来源：评论

学校读者我要写书评

暂无评论

Convolutional neural Networks for Spherical signal processing via Area-Regular Spherical Haar Tight Framelets

引用

IEEE TRANSACTIONS ON neural NETWORKS AND LEARNING SYSTEMS 2022年第4期35卷 4400-4410页

作者： Li, Jianfei Feng, Han Zhuang, Xiaosheng City Univ Hong Kong Dept Math Hong Kong Peoples R China

In this article, we develop a general theoretical framework for constructing Haar-type tight framelets on any compact set with a hierarchical partition. In particular, we construct a novel area-regular hierarchical partition on the two spheres and establish its corresponding spherical Haar tight framelets with directionality. We conclude by evaluating and illustrate the effectiveness of our area-regular spherical Haar tight framelets in several denoising experiments. Furthermore, we propose a convolutional neural network (CNN) model for spherical signal denoising, which employs fast framelet decomposition and reconstruction algorithms. Experiment results show that our proposed CNN model outperforms threshold methods and processes strong generalization and robustness.

关键词： Convolutional neural networks Convolution Noise reduction Urban areas signal representation signal denoising Robustness Area regular bounded domains convolutional neural network (CNN) directional framelets image denoising spherical Haar framelets spherical signals tight framelets

来源：评论

学校读者我要写书评

暂无评论

Blind image separation based on attentional generative adversarial network

引用

JOURNAL OF AMBIENT INTELLIGENCE AND HUMANIZED COMPUTING 2022年第3期13卷 1397-1404页

作者： Sun, Xiao Xu, Jindong Ma, Yongli Zhao, Tianyu Ou, Shifeng Peng, Lizhi Yantai Univ Sch Comp & Control Engn Yantai Peoples R China Yantai Univ Sch Optoelect Informat Sci & Technol Yantai Peoples R China Univ Jinan Prov Key Lab Network Based Intelligent Comp Jinan Peoples R China

Mixing signal separation is an important field of image processing. However, traditional blind source separation (BSS) algorithms were proposed to solve this task utilizing multiple signal constraints such as independent, non-Gaussian, low rank, sparsity, temporal continuity etc. What's more, as a case of ill-conditioned signal mixing, the single-channel blind source separation (SCBSS) is more difficult. Because neural networks have strong adaptability and self-organization capability, neural network methods based on training and learning ideas are favored by researchers. However, most BSS methods based on neural network are limited by small sample sizes. Among various neural network, generative adversarial network (GAN) has emerged as an interesting candidate because it is free from statistical constraints and samples. Therefore, we present a single-channel blind image separation algorithm based on attention mechanism GAN, coined AGAN, which uses an end-to-end manner, and it will have more hopeful prospects in the blind image separation task. The network with feature extraction, as well as edge guidance to data creates a new way to iteratively separate mixing images. The experimental results show that AGAN can effectively separate the source signal in the mixing images compared with the neural egg separation (NES) algorithm, which is a neural network separation algorithm. Compared with the classical blind source separation algorithms, this method has better separation performance.

关键词： Blind image separation Single channel Generative adversarial network Attention mechanism

来源：评论

学校读者我要写书评

暂无评论

Hierarchical multi-scale stereoscopic image quality assessment based on visual mechanism

引用

signal image AND VIDEO processing 2022年第5期16卷 1177-1185页

作者： Chang, Yongli Li, Sumei Zhao, Ping Tianjin Univ Sch Elect & Informat Engn Tianjin 300072 Peoples R China

The recent advance on stereoscopic image quality assessment (SIQA) models has been remarkably improved due to the pervasive application of convolutional neural network (CNN). Although the current CNN-based methods have achieved good results, these methods only extract single scale features at the same level. And some CNN-based methods directly take left and right images as an input of the network ignoring the visual fusion mechanism. In this work, a hierarchical multi-scale no-reference SIQA method is proposed based on dilated convolution. Multi-scale module constructed by standard convolution will lead to a sharp increase in the number of model parameters. On the contrary, the dilated convolution can restrain the increase in the number of model parameters and enlarge the receptive field. Therefore, dilated convolution is used to simulate the multi-scale characteristics of human vision. In addition, instead of left and right images, the cyclopean image generated by a new method is used as the input of the network. Experimental results on four public databases show that the proposed model is superior to the state-of-the-art SIQA methods.

关键词： Convolution neural network Cyclopean image Dilated convolution Hierarchical multi-scale Stereoscopic image quality assessment

来源：评论

学校读者我要写书评

暂无评论

ASQ-FastBM3D: An Adaptive Denoising Framework for Defending Adversarial Attacks in Machine Learning Enabled Systems

引用

IEEE TRANSACTIONS ON RELIABILITY 2023年第1期72卷 317-328页

作者： Xu, Guangquan Han, Zhengbo Gong, Lixiao Jiao, Litao Bai, Hongpeng Liu, Shaoying Zheng, Xi Huanghai Univ Sch Big Data Qingdao 266427 Peoples R China Hiroshima Univ Grad Sch Adv Sci & Engn Hiroshima 7398511 Japan Macquarie Univ Dept Comp N Ryde NSW 2109 Australia

Machine learning has made significant progress in image recognition, natural language processing, and autonomous driving. However, the generation of adversarial examples has proved that the machine learning system is unreliable. By adding imperceptible perturbations to clean images can fool the well-trained machine learning systems. To solve this problem, we propose an adaptive image denoising framework Adaptive Scalar Quantization (ASQ-FastBM3D). The ASQ-FastBM3D framework combines the ASQ method with the FastBM3D algorithm. The adaptive scalar quantization is the improvement of scalar quantization, which is used to eliminate most of the perturbations. FastBM3D is proposed to improve the quality of the quantified image. The running time of FastBM3D is 50% less than that of BM3D. Compared with some traditional filter methods and some state-of-the-art neural network methods for recovering the adversarial examples, the accuracy rate of our ASQ-FastBM3D method is 99.73% and the F1 score is 98.01%, which is the highest.

关键词： Perturbation methods neural networks Quantization (signal) Training Robustness Generative adversarial networks Deep learning Adversarial example BM3D image denoising machine learning scalar quantization

来源：评论

学校读者我要写书评

暂无评论

Matrix-based Fourier analysis of matrix signals and systems for polarization optics

引用

JOURNAL OF THE OPTICAL SOCIETY OF AMERICA A-OPTICS image SCIENCE AND VISION 2024年第10期41卷 1969-1978页

作者： Wang, Wei Heriot Watt Univ Inst Photon & Quantum Sci Sch Engn & Phys Sci Edinburgh EH14 4AS Scotland

Matrix functions are, of course, indispensable and of primary concern in polarization optics when the vector nature of light has been considered. This paper is devoted to investigating matrix-based Fourier analysis of twodimensional matrix signals and systems. With the aid of the linearity and the superposition integral of matrix functions, the theory of linear invariant matrix systems has been constructed by virtue of six matrix-based integral transformations [i.e., matrix (direct) convolution, matrix (direct) correlation, and matrix element-wise convolution/correlation]. Properties of the matrix-based Fourier transforms have been introduced with some applications including the identity impulse matrix, matrix sampling theorem, width, bandwidth and their uncertainty relation for the matrix signal, and Haagerup's inequality for matrix normalization. The coherence time and the effective spectral width of the stochastic electromagnetic wave have been discussed as an application example to demonstrate how to apply the proposed mathematical tools in analyzing polarization-dependent Fourier optics.

关键词： Fourier transforms Information processing Mueller matrices Optical transfer functions Phase space analysis methods Point spread function

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：