A5100406913- Image and Signal Denoising Methods
- Advanced Image Fusion Techniques
- Image Enhancement Techniques
- Advanced Image Processing Techniques
- Remote-Sensing Image Classification
- Sparse and Compressive Sensing Techniques
- Photoacoustic and Ultrasonic Imaging
- Advanced Adaptive Filtering Techniques
- Bayesian Methods and Mixture Models
- Antenna Design and Optimization
- Structural Health Monitoring Techniques
- Video Surveillance and Tracking Methods
- Neural Networks and Applications
- Blind Source Separation Techniques
- Computer Graphics and Visualization Techniques
- Water Quality Monitoring Technologies
- Advanced Neural Network Applications
- Advanced Computational Techniques and Applications
- Bayesian Modeling and Causal Inference
- Rough Sets and Fuzzy Logic
- Diverse Scientific and Engineering Research
- Advanced Vision and Imaging
- Simulation Techniques and Applications
- Electromagnetic Scattering and Analysis
- Matrix Theory and Algorithms
Zhejiang Ocean University
2017-2024
Guizhou Electromechanical Research and Design Institute
2022
Chinese Academy of Sciences
2017
Institute of Geochemistry
2017
Tsinghua University
2008
Xi'an Jiaotong University
2003
We introduce X-Dyna, a novel zero-shot, diffusion-based pipeline for animating single human image using facial expressions and body movements derived from driving video, that generates realistic, context-aware dynamics both the subject surrounding environment. Building on prior approaches centered pose control, X-Dyna addresses key shortcomings causing loss of dynamic details, enhancing lifelike qualities video animations. At core our approach is Dynamics-Adapter, lightweight module...
Visual reasoning refers to the task of solving questions about visual information. Current methods typically employ pre-trained vision-language model (VLM) strategies or deep neural network approaches. However, existing efforts are constrained by limited interpretability, while hindering phenomenon underspecification in question text. Additionally, absence fine-grained knowledge limits precise understanding subject behavior tasks. To address these issues, we propose VIKSER (Visual...
Regularization by denoising (RED) framework has shown impressive performance for many imaging inverse problems, leveraging the method in defining an explicit regularization. In this letter, we propose a novel SLN-RED scheme image restoration exploiting local and nonlocal denoisers simultaneously. Theoretically, proves that <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">bounded</i> denoisers, under ADMM with continuation strategy converges to...
Hyperspectral images (HSIs) are frequently contaminated by different noises (Gaussian noise, stripe deadline impulse noise) in the acquisition process as a result of observation environment and imaging system limitations, which makes image information lost difficult to recover. In this paper, we adopt 3D-based SSCA block neural network U-Net architecture for remote sensing HSI denoising, named SSCANet (Spatial Spectral-Channel Attention Network), is mainly constructed so-called block. By...
Data augmentation (DA) is an effective way to enrich the richness of data and improve a model’s generalization ability. It has been widely used in many advanced vision tasks (e.g., classification, recognition, etc.), while it can hardly be seen hyperspectral image (HSI) tasks. In this paper, we analyze whether existing methods are suitable for task HSI denoising find that biggest challenge lies neither losing spatial information original nor destroying correlation between various bands...
The second generation bandelet transform uses the two‐dimensional (2D) separable wavelet to improve its image denoising and compression performance. However, 2D is not a shift‐invariant therefore cannot capture geometric information well. authors propose hybrid method in which replaced with non‐subsampled contourlet transform. results of application proposed several greyscale colour benchmark images contaminated various levels Gaussian white noise Poisson indicate that has good peak...
This paper presents a novel video enhancement system based on an adaptive spatio-temporal connective (ASTC) noise filter and piecewise mapping function (APMF). For ill-exposed videos or those with much noise, we first introduce local image statistic to identify impulse pixels, then incorporate it into the classical bilateral form ASTC, aiming reduce mixture of most two common types noises—Gaussian noises in spatial temporal directions. After removal, enhance contrast APMF statistical...
The joint time-frequency analysis method is adopted to study the nonlinear behavior varying with instantaneous response for a class of S.D.O.F system. A masking operator, together conception effective region asymptotic signal are defined here. Based on these mathematical foundations, so-called skeleton linear model (SLM) constructed which has similar characteristics Two curves deduced can indicate stiffness and damping in relationship between SLM system, both parameters solutions, clarified....
In the previous paper, a class of nonlinear system is mapped to so-called skeleton linear model (SLM) based on joint time-frequency analysis method. Behavior may be indicated quantitatively by variance coefficients SLM versus its response. Using this we propose an identification method for systems nonstationary vibration data in paper. The key technique procedure filtering which solution extracted from response corresponding system. Two methods are discussed here. One quadratic distribution...
Object detection is a popular research field in deep learning. People usually design large-scale convolutional neural networks to continuously improve the accuracy of object detection. However, special application scenario using robot for underwater fish detection, due computational ability and storage space are limited, which leads problem low recognition fish. In this paper, an improved Ghost-YOLOv5 network based on attention mechanism proposed, use Ghostconvolution GhostNet replace...
Wavelet tight frames have been actively investigated for various image restoration problems. In this paper, we introduce an analysis-sparsity model via <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\ell _2$</tex-math></inline-formula> -relaxed truncated _0$</tex-math></inline-formula> regularization and nonlocal estimation, the resulted nonconvex minimization problem is tackled by a proximal alternating...
Current single image derain methods cannot solve the heavy rain situation well. In this paper, based on physical model of a rainy image, we build two-stage network, TSF-Net, which combines model-driven and data-driven methods. The first stage gets streaks, atmospheric light, transmission map to obtain coarse rain-free by model. second is fully convolutional neural network with structure U-Net. proposed Multi-Scale Projection Fusion Block (MSPFB) module, can perceive spatial information...
Underwater images will produce severe noise during shooting, transmission and underwater backscattering, which seriously affects the accuracy of stereo matching target positioning. For this reason, paper proposes an improved wavelet denoising method. First, a new continuous threshold function is designed to overcome problem that soft weaken image feature information hard not at point. After that, considering different propagation characteristics signal scales transform, adaptive estimation...
In this paper, an effective method of using gradient information to estimate atmospheric light is proposed, which contributes handle the problem low contrast and detail degradation in outdoor images taken under bad weather. This based on observation that visibility region with dense fog image low, lead small value. We detect foggy area by cutting expanding high corresponding place input image. Then, global can be estimated directly. Experimental results a variety hazy demonstrate efficiency...
Non-blind image deconvolution (NBID) has been a significant and long-standing research topic in the field of imaging sciences. While enduring efforts have made over past decades, they are extensively focusing on scenario where blur process is artificially assumed, e.g., typically under periodic boundary condition, as involved matrix inversion can be efficiently tackled using fast Fourier transforms (FFTs). However, despite its popularity, these conditions not always realistic assumptions...
The massive multiple-input and multiple-output (MIMO) system based on channel state information (CSI) is the core technology of next-generation communication. As complexity CSI matrix gradually increases, feedback becomes more challenging. deep learning (DL) has been successful in frequency-division duplex (FDD) MIMO systems. In this paper, we propose a complex-valued lightweight neural network MADNet for feedback. an encoder-decoder structure, which enriches extraction by adopting modules...
In massive multiple-input and multiple-output (mMIMO) systems, it is critical to acquire accurate channel state information (CSI) guarantee the efficiency robustness of wireless communications. However, sending CSI transmitters requires extra bandwidth consumption with transmission delays, especially for frequency division duplexing (FDD)scheme. According our research, a matrix has different densities in domains. Thus, this paper we propose processing deep learning framework, namely...
In this paper, a new preconditioner for numerical solutions of symmetric indefinite linear systems is presented. The called as product constructed through the two fairly simple preconditioners. eigenvalues distribution and form eigenvectors preconditioned matrix are analyzed. Numerical experiments illustrate effectiveness preconditioner.