检索结果-内蒙古大学图书馆

Robust steganography with boundary-preserving overflow alleviation and adaptive error correction

EXPERT SYSTEMS WITH APPLICATIONS 2025年 281卷

作者： Cheng, Yu Luo, Zhenlin Yin, Zhaoxia East China Normal Univ Shanghai Key Lab Multidimens Informat Proc Shanghai 200241 Peoples R China

With the rapid evolution of the Internet, the vast amount of data has created opportunities for fostering the development of steganographic techniques. However, traditional steganography faces challenges in social networks due to lossy operations such as spatial truncation during JPEG recompression, with limited research on their effects. Existing methods aim to ensure the stability of the quantized coefficients by reducing the effects of spatial truncation. Nevertheless, these approaches may induce notable alterations to image pixels, potentially compromising anti-steganalysis performance. In this study, we analyze the overflow characteristics of spatial blocks and observe that pixel values at the boundaries of spatial blocks are more prone to overflow. Based on this observation, we propose a preprocessing method that performs overflow removal operations according to the actual overflow conditions of spatial blocks, enhancing coefficient stability while minimizing modifications to spatial block boundaries, thereby ensuring image quality. Subsequently, we employ adaptive error correction coding to reduce coding redundancy, thereby augmenting robustness and mitigating its impact on anti-steganalysis performance. The experimental results indicate that the proposed method possesses a strong embedding capacity, maintaining a high level of robustness while enhancing security.

关键词： Robust steganography Overflow Adaptive error correction Dither modulation Social networks

来源：评论

学校读者我要写书评

暂无评论

MAED: Mask assignment encoder decoder solver for multiple patterning layout decomposition

引用

EXPERT SYSTEMS WITH APPLICATIONS 2025年 268卷

作者： Yu, Fang Shen, Jiwei Lyu, Shujing Lu, Yue East China Normal Univ Shanghai Key Lab Multidimens Informat Proc Shanghai 200241 Peoples R China

Multiple patterning layout decomposition (MPLD) plays a vital role in multiple patterning lithography (MPL) which assigns features in a layout into multiple physical masks. Existing methods, such as integer linear programming (ILP), manually design sophisticated formulations which can find the optimal solutions but exhibit decreased runtime performance for dense layouts. Besides, the need for problem specific designs cannot guarantee the trivial extension to other conditions. Regarding the issues above, in this paper, we revisit the objective of MPLD and design a general-purpose formulation for learning methods that is broadly applicable to layout decomposition problems. Moreover, a graph learning framework - MAED is introduced to solve the proposed formulation. Through minimizing the carefully designed differentiable objective, MAED can be trained unsupervised purely on small scale layouts and applied quickly to arbitrarily large ones via message passing mechanism. Combined with the greedy based color refinement hammer, we further improve the local quality of solutions. Experimental results reveal that our method can achieve more than 100x speedup compared to ILP and displays better quality-speed trade-offs compared with other approximation solvers.

关键词： Graph neural network Design for manufacturability Layout decomposition VLSI design

来源：评论

学校读者我要写书评

暂无评论

GCENet: A geometric correspondence estimation network for tracking and loop detection in visual-inertial SLAM

引用

EXPERT SYSTEMS WITH APPLICATIONS 2025年 262卷

作者： Zhou, Jichao Shen, Jiwei Lyu, Shujing Lu, Yue East China Normal Univ Shanghai Key Lab Multidimens Informat Proc Shanghai 200241 Peoples R China

Establishing robust and effective data correlation has been one of the core problems in visual based SLAM (Simultaneous Localization and Mapping). In this paper, we propose a geometric correspondence estimation network, GCENet, tailored for visual tracking and loop detection in visual-inertial SLAM. GCENet considers both local and global correlation in frames, enabling deep feature matching in scenarios involving noticeable displacement. Building upon this, we introduce a tightly-coupled visual-inertial state estimation system. To address challenges in extreme environments, such as strong illumination and weak texture, where manual feature matching tends to fail, a compensatory deep optical flow tracker is incorporated into our system. In such cases, our approach utilizes GCENet for dense optical flow tracking, replacing manual pipelines to conduct visual tracking. Furthermore, a deep loop detector based on GCENet is constructed, which utilizes estimated flow to represent scene similarity. Spatial consistency discrimination on candidate loops is conducted with GCENet to establish long-term data association, effectively suppressing false negatives and false positives in loop closure. Dedicated experiments are conducted in EuRoC drone, TUM-4Seasons and private robot datasets to evaluate the proposed method. The results demonstrate that our system exhibits superior robustness and accuracy in extreme environments compared to the state-of-the-art methods.

关键词： Visual-inertial SLAM Geometric correspondence estimation Visual tracking Loop closure detection

来源：评论

学校读者我要写书评

暂无评论

Combining SEFDM Technique With IBFD Communication System in an ROF Link for Achieving Ultra-High Spectral Efficiency

引用

JOURNAL OF LIGHTWAVE TECHNOLOGY 2025年第3期43卷 1089-1097页

作者： Chen, Yu Chen, Yang East China Normal Univ Sch Commun & Elect Engn Shanghai Key Lab Multidimens Informat Proc Shanghai 200241 Peoples R China

Oriented to the requirements of future high-frequency wireless communication systems, this paper proposes a method of combining spectrally efficient frequency-division multiplexing (SEFDM) technique with the in-band full-duplex (IBFD) communication system and applying it to microwave photonic transmission links to achieve ultra-high spectral efficiency. The microwave photonic link not only realizes the transmission of SEFDM signals received by the IBFD system but also realizes the analog self-interference cancellation (SIC) function. The digital SIC and SEFDM demodulation functions are seamlessly integrated following the analog SIC. Utilizing a cross-iterative algorithm, the adverse effects of the signal of interest (SOI) on the digital SIC of the received signal are mitigated, leading to a substantial improvement in both the SIC depth and demodulation performance of the SEFDM signal compared to the conventional least squares (LS) algorithm. An experiment has been conducted. The SOI and self-interference (SI) have symbol rates of 300 Msym/s and 600 Msym/s, respectively, and share a center frequency of 2 GHz and a bandwidth compression factor of 0.8. When the SI to SOI power ratio is 10.3 dB, the analog SIC depth is around 18.3 dB. The conventional LS algorithm achieves a digital SIC depth of 12.6 dB but the error vector magnitude (EVM) is 13.2%. In contrast, our proposed iterative method improves the SIC depth to 15.1 dB and reduces the EVM to 4.1%. The feasibility of the system is also verified by incorporating a 25.2-km fiber.

关键词： Interference cancellation Demodulation Microwave photonics Optical fibers Microwave theory and techniques Microwave communication Optical fiber dispersion Optical fiber communication Communication systems Signal processing algorithms Digital signal processing in-band full-duplex microwave photonics self-interference cancellation (SIC) spectrally efficient frequency-division multiplexing (SEFDM)

来源：评论

学校读者我要写书评

暂无评论

RefineStyle: Dynamic Convolution Refinement for StyleGAN 7th

RefineStyle: Dynamic Convolution Refinement for StyleGAN

引用

7th Chinese Conference on Pattern Recognition and Computer Vision

作者： Xia, Siwei Hu, Xueqi Sun, Li Li, Qingli East China Normal Univ Shanghai Key Lab Multidimens Informat Proc Shanghai Peoples R China

ISBN: (纸本)9789819786916;9789819786923

In StyleGAN, convolution kernels are shaped by both static parameters shared across images and dynamic modulation factors w(+) is an element of W+ specific to each image. Therefore, W+ space is often used for image inversion and editing. However, pre-trained model struggles with synthesizing out-of-domain images due to the limited capabilities of W+ and its resultant kernels, necessitating full fine-tuning or adaptation through a complex hypernetwork. This paper proposes an efficient refining strategy for dynamic kernels. The key idea is to modify kernels by low-rank residuals, learned from input image or domain guidance. These residuals are generated by matrix multiplication between two sets of tokens with the same number, which controls the complexity. We validate the refining scheme in image inversion and domain adaptation. In the former task, we design grouped transformer blocks to learn these token sets by one- or two-stage training. In the latter task, token sets are directly optimized to support synthesis in the target domain while preserving original content. Extensive experiments show that our method achieves low distortions for image inversion and high quality for out-of-domain editing.

关键词： Computer vision Generative models GAN inversion Domain adaptation

来源：评论

学校读者我要写书评

暂无评论

Comprehensive Electrothermal Characterization Analysis for Scaled Nanochannels in Gate-All-Around Field-Effect Transistors

引用

INTERNATIONAL JOURNAL OF NUMERICAL MODELLING-ELECTRONIC NETWORKS DEVICES AND FIELDS 2025年第1期38卷

作者： Wang, Ziping Li, Fei Sun, Yabin Shi, Yanling Li, Xiaoji East China Normal Univ Dept Elect Engn Shanghai Key Lab Multidimens Informat Proc Shanghai Peoples R China Empyrean Technol Co Ltd Beijing Peoples R China

Gate-all-around field-effect transistors (GAAFETs) have garnered extensive research interest and industrial attention due to the higher gate control capability and remarkable scalability. However, as the nanochannel scales down, the phonon-boundary scattering inside channels is dramatically strengthened, resulting in a significant decrease in phonon mean free path (MFP), which in turn leads to a decrease in thermal conductivity and deteriorates electrothermal characteristics. In this paper, to accurately evaluate the degradation of thermal conductivity for confined nanochannels, an analytical model is developed by revising the boundary-induced reduction function related to both nanochannel width and thickness. The results show that the thermal conductivity calculated by the proposed model agrees well with the experimental data within 1% error over large temperature range for nanosheet and nanowire structures. Moreover, significant deviations of 6.11% in on-state current and 41.7 K in temperature are observed between the proposed and conventional models for three-stacked GAAFETs. The proposed revised methodology offers invaluable insights for assessing the electrothermal characteristics of nanodevices.

关键词： nanosheet gate-all-around FET phonon-boundary scattering self-heating effects TCAD thermal conductivity model

来源：评论

学校读者我要写书评

暂无评论

Stomatopods compound eye structure-inspired circular polarization quantum well infrared photodetectors

引用

APPLIED PHYSICS B-LASERS AND OPTICS 2025年第2期131卷 1-9页

作者： Feng, Jianlin Jiang, Hengrui Zhao, Jun Xiong, Dayuan East China Normal Univ Dept Elect Key Lab Polarized Mat & Devices MOE Shanghai 200241 Peoples R China East China Normal Univ Shanghai Key Lab Multidimens Informat Proc Shanghai 200241 Peoples R China

Polarization detection is a crucial technology widely employed in infrared detection, image recognition, and space optical communication. Within the realm of infrared detection, it serves to enhance the detection of faint targets against complex backgrounds. Consequently, enhancing the performance of polarization devices has become one of the pivotal directions in quantum well infrared photodetectors (QWIPs) research. This paper proposes a novel circular polarization QWIPs inspired by stomatopods compound eye structure. By emulating the compound eye structure of the mantis shrimp, we integrate a top dielectric metasurface and a bottom one-dimensional metallic grating on QWIPs to achieve circular polarization detection. The dielectric metasurface, fabricated by etching heavily doped GaAs material, primarily functions as a quarter-wave plate, decomposing incident circularly polarized light into two orthogonally polarized linear components. The one-dimensional metallic grating at the bottom of device, acting as a linear polarizer, selectively excites surface plasmon polariton (SPP) modes, thereby distinguishing left-handed circularly polarized (LCP) light from right-handed circularly polarized (RCP) light. Finite-difference time-domain method was employed to calculate the total absorption spectrum of this integrated device. Under the LCP light incidence, the device exhibits a total absorption of 0.9 at the peak response wavelength of 7.8 mu m, with a coupling efficiency of 3553%. Conversely, under RCP light incidence, the device demonstrates a total absorption of 0.1 at the same peak response wavelength, achieving a coupling efficiency of 95%. Furthermore, the device's circular polarization extinction ratio reaches 37.4. This structure, achievable through quantum well focal plane array technology, represents a positive performing design for circular polarization QWIPs, contributing to enhancing target identification capabilities.

关键词： Quantum well infrared photodetectors

来源：评论

学校读者我要写书评

暂无评论

A method based on hybrid cross-multiscale spectral-spatial transformer network for hyperspectral and multispectral image fusion

引用

EXPERT SYSTEMS WITH APPLICATIONS 2025年 263卷

作者： Chen, Yingxia Wei, Mingming Chen, Yan Yangtze Univ Sch Comp Sci Jingzhou 430023 Peoples R China East China Normal Univ Shanghai Key Lab Multidimens Informat Proc Shanghai 200241 Peoples R China

Convolutional neural networks (CNNs) have made a significant contribution to hyperspectral image (HSI) generation. However, capturing long-range dependencies can be challenging with CNNs due to the limitations of their local receptive fields, which can lead to distortions in fused images. Transformers excel at capturing longrange dependencies but have limited capacity for handling fine details. Additionally, prior work has often overlooked the extraction of global features during the image preprocessing stage, resulting in the potential loss of fine details. To address these issues, we propose a hybrid cross-multiscale spectral-spatial Transformer (HCMSST) that combines the advantages of CNNs in feature extraction and Transformers in capturing long-range dependencies. To fully extract and retain local and global information in the shallow feature extraction phase, the network incorporates CNNs with a staggered cascade-dense residual block (SCDRB). This block employs staggered residuals to establish direct connections both within and between branches and integrates attention modules to enhance the response to important features. This approach facilitates unrestricted information exchange and fosters deeper feature representations. To address the limitations of Transformer in processing fine details, we introduce multiscale spatial-spectral coding-decoding structures to obtain comprehensive spatialspectral features, which are utilized to capture the long-range dependencies via the cross-multiscale spectralspatial Transformer (CMSST). Further, the CMSST incorporates a cross-level dual-stream feature interaction strategy that integrates spatial and spectral features from different levels and then feeds the fused features back to their corresponding branches for information interaction. Experimental results indicate that the proposed HCMSST achieves superior performance compared to many state-of-the-art (SOTA) methods. Specifically, HCMSST reduces the ERGAS metric by 3.0

关键词： Hyperspectral image (HSI) Multispectral image (MSI) Image fusion Staggered cascade-dense residual block (SCDRB) Cross-multiscale spectral-spatial Transformer (CMSST) Cross-level dual-stream feature interaction strategy

来源：评论

学校读者我要写书评

暂无评论

FAN: Fourier Assignment Network for autofocus based on deep learning

引用

OPTICS AND LASER TECHNOLOGY 2025年 186卷

作者： Liu, Qizheng Mao, Xintian Wang, Jiansheng Zhang, Qing Wang, Yan Pang, Baochuan Li, Qingli East China Normal Univ Shanghai Key Lab Multidimens Informat Proc Dongchuan Rd Shanghai Peoples R China Landing Med 818 Gaoxin Rd Wuhan Hubei Peoples R China

AI-driven digital pathology is transforming the field, prioritizing the enhancement of image quality. In the data acquisition phase, the performance of autofocus algorithms significantly influences imaging quality. To tackle this challenge, we introduce the Fourier Assignment Network (FAN), a novel and lightweight neural network architecture aimed at predicting the defocus distance in microscopic images. FAN estimates the defocus distance using single-shot images, thus eliminating the requirement for extra hardware. Extensive experiments conducted on our dataset confirm its effectiveness, with an average error of 0.9 mu m across 1374 sample groups, markedly surpassing the performance of existing lightweight neural networks. Moreover, FAN is the most efficient among surveyed networks, with only 0.14G FLOPs and 0.85M parameters. The real-time and precise autofocus capability of FAN marks a notable advancement in digital pathology, providing a dependable, hardware-independent solution to enhance diagnostic accuracy.

关键词： Autofocus Microscopy Deep learning Digital pathology

来源：评论

学校读者我要写书评

暂无评论

Mode Switchable Quantum Well Infrared Detector with Rectangular Plasmonic Microcavity Structure

引用

PLASMONICS 2025年 1-10页

作者： Zhao, Jun You, Tongtong Jiang, Hengrui Xiong, Dayuan East China Normal Univ Dept Elect Key Lab Polarized Mat & Devices MOE Shanghai 200241 Peoples R China East China Normal Univ Shanghai Key Lab Multidimens Informat Proc Shanghai 200241 Peoples R China

In this paper, a GaAs/AlGaAs quantum well infrared photodetector (QWIP) using a rectangular plasmonic microcavity structure is proposed. Based on the fact that different side lengths of the rectangular patch responding to different TM waves during surface plasmon resonance, the device has three modes of operation. In the case of TM10 and TM01 waves being incident alone, the response peaks of the device are located at 6.48 mu m and 7.38 mu m, which correspond to the two single-peak modes of Ex and Ey polarisation, respectively. The two response wavelengths generated by the two polarisation modes are independently regulated within a certain range and are not affected by each other. In the case of simultaneous incidence of TM10 and TM01 waves, the full width at half maximum (FWHM) of the device's responsivity spectrum reaches 1.56 mu m, which is broadband response mode. The simulation study reveals that the average electric field enhancement |E-Z|/|E-0| in all three modes is more than 5 times, while the single-peak and broadband response mode responsivities reach 3 A/W and 2 A/W, respectively. Our findings indicate that the average energy inside the microcavity is positively correlated with the microcavity volume within a certain range. Moreover, the coupling structure exhibits favourable incident angle compatibility, and the device performance in each mode is insensitive to the incident angle in the range of similar to 45 degrees.

关键词： Quantum well infrared detector Plasmonic microcavity Three modes of operation Broadband response

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：