检索结果-内蒙古大学图书馆

conference Applications of Digital image processing XLIV

作者： Testolina, Michela Upenik, Evgeniy Ebrahimi, Touradj Ecole Polytech Fed Lausanne EPFL Multimedia Signal Proc Grp MMSPG CH-1015 Lausanne Switzerland

ISBN: (数字)9781510645233

ISBN: (纸本)9781510645233;9781510645226

In recent years, learning-based image compression has demonstrated similar or superior performance when compared to conventional approaches in terms of compression efficiency and visual quality. Typically, learning-based image compression takes advantage of autoencoders, which are architectures consisting of two main parts: a multi-layer neural network encoder and its dual decoder. The encoder maps the input image represented in the pixel domain to a compact representation, also known as latent space. Consequently, the decoder reconstructs the original image in the pixel domain from its latent representation, as accurately as possible. Traditionally, image processing algorithms, and in particular image denoising, are applied to images in the pixel domain before compression, and eventually in some cases as a post-processing stage after decompression. In this context, the combination of denoising operations with the autoencoder might reduce the computational cost while achieving similar performance in accuracy. In this paper, the idea of combining the image denoising task with compression is examined. In particular, the integration of denoising convolutional layers in the decoder of a learning-based compression network is investigated. Results show that, while the rate-distortion performance of the method is slightly reduced, a gain in the computational complexity can be achieved.

关键词： image denoising learning-based compression compressed domain denoising latent space image processing

来源：评论

学校读者我要写书评

暂无评论

CLIFS: CLIP-DRIVEN FEW-SHOT LEARNING FOR BAGGAGE THREAT CLASSIFICATION 31

CLIFS: CLIP-DRIVEN FEW-SHOT LEARNING FOR BAGGAGE THREAT CLAS...

引用

2024 international conference on image processing

作者： Ahmed, Abdelfatah Velayudhan, Divya ElMezain, Mahmoud Al Radi, Muaz Boudiaf, Abderrahmene Hassan, Taimur Deriche, Mohamed Bennamoun, Mohammed Werghi, Naoufel Khalifa Univ Abu Dhabi U Arab Emirates Abu Dhabi Univ Abu Dhabi U Arab Emirates Ajman Univ Ajman U Arab Emirates Univ Western Australia Perth Australia

ISBN: (纸本)9798350349405;9798350349399

Baggage screening in airports is a cornerstone in airport security measures. The advent of computer vision technologies in recent years has led to the development of several automated systems for identifying security threats in baggage scans. However, existing methods struggle to adapt to new threat categories when faced with a scarcity of data samples, and the rapid emergence of new threats. Hence, in this paper, we propose a novel CLIP-driven few-shot framework (CLIFS) to explore the potential of multi-modality using text-image fusion through contrastive learning to learn relevant contextual features for recognizing security threats with limited samples. By integrating features from GPT-4 generated captions with image features, CLIFS leverages both visual and textual data to significantly improve threat classification performance with limited samples in a few-shot learning context. Our proposed CLIFS was rigorously tested on the SIXray public available baggage X-ray dataset, where it outperformed state-of-the-art by 31.3% in accuracy and 28.40% in F1-score for the challenging 5-shots scenario, demonstrating its robustness and effectiveness in classifying threats from limited data samples.

关键词： Vision-Langugae Model Contrastive Loss Few-shot learning Contrastive Language image Pretraining Baggage threat classification

来源：评论

学校读者我要写书评

暂无评论

JOINT PROBABILITY DISTRIBUTION REGRESSION FOR image CROPPING 30

JOINT PROBABILITY DISTRIBUTION REGRESSION FOR IMAGE CROPPING

引用

30th IEEE international conference on image processing (ICIP)

作者： Shi, Tengfei Chen, Chenglizhao He, Yuanbo Song, Wenfeng Hao, Aimin Beihang Univ Beijing Peoples R China China Univ Petr East China Qingdao Peoples R China Qingdao Res Inst Qingdao Peoples R China Peng Cheng Lab Shenzhen Peoples R China Informat Sci & Technol Univ Beijing Peoples R China

ISBN: (纸本)9781728198354

image cropping aims at locating a candidate (rectangle region) with the highest aesthetic quality in professional photography. One solution of the previous methods is to generate a large number of candidates and then filter them, which leads to low efficiency. Another idea directly regresses the candidate coordinates to speed up but ignores the aesthetic subjectivity of the candidate's evaluation, limiting the model's performance. In this paper, we present an Aesthetic and Composition joint Probability Distribution regression Network (ACPD-Net) to explicitly investigate the process of generating the candidate with a joint probability distribution paradigm to improve the performance of cropping results in an efficient way. The joint probability distribution paradigm between location and size branch can identify the subjective aesthetic region and satisfy the objective composition rules in an end-to-end manner. Our method has been tested on the FCDB and FLMS datasets, which shows the superiority of ACPD-Net. The code is available at https://***/flyingbird93/ACPD-Net.

关键词： image cropping image aesthetics image composition deep learning computer vision

来源：评论

学校读者我要写书评

暂无评论

A real-time image-based rendering framework

A real-time image-based rendering framework

引用

1st international conference on 3DTV

作者： Weigel, Christian Werner, Stefan Schuebel, Peter Tech Univ Ilmenau Inst Media Technol D-98684 Ilmenau Germany

ISBN: (纸本)9781424407217

We present a software framework designed for real-time image-based rendering applications. Due to its modular structure the framework suits for a number of tasks mainly in the computer vision area. First, we introduce the state of the art software design using parallel processing paradigms. Then, we show examples of 3DTV applications based on the system namely the preprocessing and synthesis of so called 3D video objects based on the theory of the trifocal transfer. We point out that due to the possibility of freely configuring the system the calculation of pre-processing and synthesis algorithms are accelerated significantly. Finally, we present possible future applications.

关键词： IBR 3D video object trifocal transfer real-time processing multithreading

来源：评论

学校读者我要写书评

暂无评论

Color Light Field image Encryption Algorithm Based on DNA Coding and Self-Controlled Chaotic System

Color Light Field Image Encryption Algorithm Based on DNA Co...

引用

2021 international conference on signal processing and Communication Technology, SPCT 2021

作者： Zhang, Xinyue Bai, Enjian Jiang, Xue-Qin Wu, Yun School of Information Science & Technology Donghua University Shanghai China

ISBN: (数字)9781510653344

ISBN: (纸本)9781510653337

Color light field images encryption algorithm based on frequency domain DNA coding and integer wavelet transform (IWT) is introduced. First, the light field image is decomposed into sub-view images in the angular domain plane. Subsequently, a low-frequency coefficient matrix and three high-frequency coefficient matrices are obtained in the frequency domain using the two-dimensional IWT. The low-frequency coefficient matrix is encrypted using DNA coding and a self-controlled chaotic system based on PWLCM. The high-frequency coefficient matrix is scrambled, and then the reconstructed intermediate encrypted image is obtained by a two-dimensional integer wavelet inverse transform. Finally, in the spatial domain, a diffusion operation is performed to obtain the final encrypted image. The simulation shows that this algorithm has the advantages of large key space, high operation efficiency, good encryption security and strong resistance to various attacks. © 2022 SPIE.

关键词： Frequency domain analysis

来源：评论

学校读者我要写书评

暂无评论

DBSwin: Transformer based dual branch network for single image deraining

引用

JOURNAL OF INTELLIGENT & FUZZY SYSTEMS 2022年第4期43卷 5109-5123页

作者： Tan, Fuxiang Qian, Yurong Kong, Yuting Zhang, Hao Zhou, Daxin Fan, Yingying Chen, Long Xiao, Zhengqing Xinjiang Univ Sch Software Urumqi Peoples R China Xinjiang Univ Key Lab Software Engn Urumqi Peoples R China Key Lab Signal Detect & Proc Xinjiang Uygur Auton Urumqi Peoples R China

Rain streaks severely affect the perception of the content and structure of an image so high-performance deraining algorithms are needed in order to eliminate the effects of various rain streaks for high-level computer vision tasks. Although much progress has been made with existing deraining methods, the task of single image deraining remains challenging. In this paper, we first point out that existing Transformers lack sufficient ability to capture channel attention which restricted the ability of models in deraining. To improve the performance of deraining model, we propose a dual branch deraining network based on Transformer. One branch uses dense connections to connect Transformer modules which embed the attention of a composite channel. This branch captures channel attention more finely to learn the representation of rain streaks features. The other branch first obtains features at different scales by gradually expanding the receptive field, then uses these features to obtain attention for regional features, and finally uses the attention to guide the model to focus on areas of high rain streaks density and large scales. By fusing these two branches, the model is able to capture channel attention more finely and to focus on regions of high rain streaks density and large scales. The extensive experimental results on synthetic and real datasets demonstrate that the proposed method outperforms most advanced deraining methods.

关键词： Deraining transformer deep learning image processing channel attention

来源：评论

学校读者我要写书评

暂无评论

RW-HAZE: A REAL-WORLD BENCHMARK DATASET TO EVALUATE QUANTITATIVELY DEHAZING algorithms 29

RW-HAZE: A REAL-WORLD BENCHMARK DATASET TO EVALUATE QUANTITA...

引用

IEEE international conference on image processing (ICIP)

作者： Chen, Jiyou Wang, Shengchun Liu, Xin Yang, Gaobo Hunan Univ Sch Comp Sci & Elect Engn Changsha 410082 Peoples R China Hengyang Normal Univ Sch Phys & Elect Engn Hengyang 421008 Peoples R China Hunan Normal Univ Sch Informat Sci & Engn Changsha 410081 Peoples R China Hunan Appl Technol Univ Changde 415100 Peoples R China

ISBN: (数字)9781665496209

ISBN: (纸本)9781665496209

Most existing image dehazing approaches are able to achieve desirable results whose differences are too subtle for people to qualitatively judge. Therefore, it is important to adopt quantitative assessment on real-world hazy images. However, many dehazing works have not been quantitatively evaluated on real-world hazy images due to the lack of appropriate real-world datasets. In this work, we attempt to address the issue and present a well-aligned real-world benchmark dataset, namely RW-Haze, for image dehazing evaluation, which had been lacking for a long period of time. It contains 210 pairs of well-aligned haze-free images and hazy images with distinct haze densities, which were captured from six cities in China by fixed cameras. To the best of our knowledge, RW-Haze is the first real-world dataset that is made up of well-aligned image pairs of haze-free and hazy images with diverse haze levels. We select 13 state-of-the-art single image dehazing works for making comprehensive evaluations among them on RW-Haze dataset. Experimental results show that there still exist rich rooms for image dehazing research to improve its robustness on natural hazy images, especially dense haze scenes.

关键词： image dehazing benchmark dataset performance evaluation

来源：评论

学校读者我要写书评

暂无评论

image super-resolution method based on generative adversarial network 4

Image super-resolution method based on generative adversaria...

引用

4th international conference on advanced Electronic Materials, Computers and Software Engineering, AEMCSE 2021

作者： Lijun, Yao Xiaoming, Zhu Fan, Luo Gang, Sun Zhou, Chen Jing, Yan Min, Zhou Yongchang, Cai Lingling, Wang Zelong, Cao Lan, Peng Fengqing, Bai Zifang, You Hongqiu, Xiao Pengjian, Li Zhisheng, Lv Qiang, Tang Changzhutan Tobacco Logistics Co.Ltd Ltd. of Hunan Province ChangSha410000 China Hunan Tobacco Company Chagnsha410000 China Hunan Tobacco Workers Training Center Xiangtan411100 China Hunan Normal University College of Information Science and Engineering ChangSha410081 China Hunan Normal University College of Engineering and Design ChangSha410081 China

ISBN: (纸本)9781665415965

Although deep convolutional neural networks have made breakthroughs in the accuracy and speed of single-image super-resolution, there are still many unsolved problems: firstly, How to refine the texture problem when performing super-resolution processing at a larger magnification ratio. Secondly, the existing convolutional neural network image super-resolution algorithms are prone to overfitting and insufficient convergence of the loss function. Aiming at two problems, an image super-resolution method based on generative adversarial network is proposed. The feature map is spatially transformed on the network to solve the problem of refined texture, combined with CycleGAN and SRGAN, the network structure is improved and the loss function is optimized, and the SRCICGAN algorithm is proposed to restore the four times down-sampled image to solve the loss function problem. The experiment is compared with the latest six methods on three data sets. The PSNR and SSIM indicators are 1.92% and 5.49% higher in the Flickr2K data set, respectively, and better visual effects can be obtained in terms of detailed texture. © 2021 IEEE.

关键词： Generative adversarial networks

来源：评论

学校读者我要写书评

暂无评论

A Multiscale Dual-Attention based Convolutional Neural Network for Ship Classification in SAR image

A Multiscale Dual-Attention based Convolutional Neural Netwo...

引用

2021 IEEE international conference on signal processing, Communications and Computing, ICSPCC 2021

作者： Zhou, Gaoyu Zhang, Gong Fang, Zheng Dai, Qijun Ministry of Education Nanjing University of Aeronautics and Astronautics Key Lab of Radar Imaging and Microwave Photonics Nanjing China

ISBN: (纸本)9781665429184

Classical ship target classification algorithms in synthetic aperture radar (SAR) images mainly rely on manual extraction of simple physical features to distinguish different ship types. A novel multiscale dual-Attention (MDA) based convolutional neural network (CNN) is proposed in this paper to highlight the deep distinguishable features extracted from SAR images for classification. First, we apply the data enhancement method to expand the limited labeled sample set to alleviate the defect of unbalanced dataset. Second, the squeeze-And-excitation(SE) block is adopted to learn the initial weight of the feature maps to mitigate the negative effects of large similarity between different ship classes. Third, the weighted feature maps are sent into two parallel self-Attention blocks, and the semantic dependence on spatial dimension and channel dimension is modeled respectively. Feature fusion is carried out to further improve feature discrimination ability. The results of ablation experiments and comparative experiments on the open SAR ship dataset show that this method can mine deep discriminant features effectively, and further improve the classification accuracy of ships in SAR images compared with baseline method and other latest methods. © 2021 IEEE.

关键词： Ships

来源：评论

学校读者我要写书评

暂无评论

Analysis and Application of Multispectral image processing Techniques Applied to Soybean Crops from Drones Vision System 18

Analysis and Application of Multispectral Image Processing T...

引用

18th international conference on Informatics in Control, Automation and Robotics (ICINCO)

作者： Gonzalez, Evelio Nunez, Cristhian Salinas, Jose Rodas, Jorge Rodas, Mariela Paiva, Enrique Kali, Yassine Saad, Maarouf Lesme, Fernando Lesme, Jose Gonzalez, Luis Maldonado, Belen Rodriguez-Pineiro, Jose Univ Catolica Nuestra Senor Asuncion Unidad Pedag Caacupe Caacupe Paraguay TECHA Caacupe Paraguay Univ Nacl Asuncion Fac Ingn Lab Power & Control Syst Luque Paraguay Inst Paraguayo Tecnol Agr Caacupe Paraguay Ecole Technol Super GREPCI Lab Montreal PQ Canada Tongji Univ Coll Elect & Informat Engn Shanghai Peoples R China

ISBN: (纸本)9789897585227

Drones are important in precision agriculture applications since they represent a new tool that can increase crop production. In this context, the digital processing of the images obtained from multispectral cameras integrated into the drones makes it possible to analyze the stress state of the crops, their vigor, a burned area, among others. The latter are usually obtained through proprietary applications with very high subscription costs. For this reason, this article presents the step-by-step implementation process of the different methods or algorithms to be applied to multispectral images using the open-source Python programming language. We use a soybean crop as an example of the application, and the results obtained from applying the digital image processing algorithms are presented.

关键词： Drones Multispectral Imaging Digital signal processing Precision Agriculture

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：