A5102825993- Advanced Vision and Imaging
- Advanced Image Processing Techniques
- GaN-based semiconductor devices and materials
- Image and Signal Denoising Methods
- Video Coding and Compression Technologies
- Semiconductor Quantum Structures and Devices
- Photonic and Optical Devices
- Ga2O3 and related materials
- Advanced Data Compression Techniques
- Gas Sensing Nanomaterials and Sensors
- Thin-Film Transistor Technologies
- Microwave Engineering and Waveguides
- Industrial Gas Emission Control
- Medical Image Segmentation Techniques
- Advanced Image and Video Retrieval Techniques
- Antifungal resistance and susceptibility
- Advancements in Solid Oxide Fuel Cells
- Robotic Locomotion and Control
- Radiation Detection and Scintillator Technologies
- Sustainable Industrial Ecology
- Visual perception and processing mechanisms
- Economic and Technological Systems Analysis
- EEG and Brain-Computer Interfaces
- Fungal Infections and Studies
- Chemical and Physical Properties of Materials
Wuhan Ship Development & Design Institute
2024
Sun Yat-sen University
2008-2023
Xidian University
2017-2018
Fudan University
2008-2013
Zhongshan Hospital
2008
The First Affiliated Hospital, Sun Yat-sen University
2008
Kunming Metallurgy College
2007
Memorial University of Newfoundland
2003
We present a novel image interpolation method based on variational models with both smoothing and orientation constraints. By introducing the constraint, we simplify nonlinear PDE problem into series of problems explicit solutions. In our model, gradient directions for interpolated pixels are first estimated using modified diffusion method. Using these adaptive directional is carried out. An effective numerical implementation presented case upsampling by factors two. This had very low...
The transmission of virtual reality (VR) videos requires huge bandwidth, which brings great challenges for the system to implement real-time applications. This letter proposes a scalable full-panorama video coding method adapt insufficient in user's movement information is utilized as feedback from VR device encoder. regions that user interested are first mapped and then coded high quality, while others low quality. different-quality achieved through efficiency (SHVC), only makes limited...
We present a simple but effective technique to smooth out textures while preserving the prominent structures. Our method is built upon key observation---the coarsest level in Gaussian pyramid often naturally eliminates and summarizes main image This inspires our central idea for texture filtering, which progressively upsample very low-resolution full-resolution smoothing result with well-preserved structures, under guidance of each fine-scale its associated Laplacian level. show that...
Polarization induced negative charges at pseudo-crystalline InGaN/GaN heterointerfaces were utilized to increase the interface potential and form a bidirectional barrier, leading an ultrasensitive photoresponse.
Visible-light detection with high sensitivity and strong wavelength selectivity is highly desired in emerging applications. Here, we demonstrate a high-performance visible-light photodetector an active region composed of polarization induced barrier single-carrier superlattices (SCSLs). The at SCSLs/GaN heterointerface brings both low dark current gain originating from the photoinduced reduction effect. Meanwhile, designed InGaN/GaN SCSLs allow photoelectrons quantum wells to escape, but...
Abstract Liquid Scintillation Counting (LSC) is a technique widely employed for measuring the activity of radioactive isotopes, finding applications in fields such as isotope measurements, nuclear material analysis, environmental monitoring, and medicine. This study based on Triple-to-Double Coincidence Ratio (TDCR) method within liquid scintillation counting, proposing novel TDCR-LSC detector structure with TriL-shaped optical chamber. Compared to traditional TDCR structures, chamber...
We propose a practical downsampling compression scheme, as shown in Fig 1, to use neural network pre-processing module for traditional codecs improve performance. Specifically, we network-based method video performance by high-resolution videos. Our method, called the rate-guided arbitrary rescaling (RARN), uses pre-trained entropy [1] estimate bitrate and guide sampling process. also transformer-based virtual codec (TVC) simulate of codecs, using swing-transformer-based invertible learn...
In this paper, we present a novel PDE based error concealment algorithm. We formulate the problem as sequential optimization with both smoothing and orientation constraints. By introducing constraint convert nonlinear variational into that is well posed which can be solved without iterative operations. A modified diffusion scheme presented able to reconstruct complex patterns within blocks have been lost in an image. intensity reconstruction stage follows diffusion, performed on estimates...
Most video platforms provide streaming services with different qualities, and the quality of is usually adjusted by resolution videos. So high-resolution videos need to be downsampled for compression. In order solve problem coding at resolutions, we propose a rate-guided arbitrary rescaling network (RARN) resizing before encoding. To help RARN compatible standard codecs generate compression-friendly results, an iteratively optimized transformer-based virtual codec (TVC) introduced simulate...
Channel prediction is critical to address the channel aging issue in mobile scenarios. Existing techniques are mainly designed for discrete prediction, which can only predict future a fixed time slot per frame, while other intra-frame channels usually recovered by interpolation. However, these approaches suffer from serious interpolation loss, especially millimeter wave communications. To solve this challenging problem, we propose tensor neural ordinary differential equation (TN-ODE) based...