A5008612863- Advanced Image Processing Techniques
- Advanced Vision and Imaging
- Image and Signal Denoising Methods
- Image Enhancement Techniques
- Image Processing Techniques and Applications
- Cell Image Analysis Techniques
- Optical measurement and interference techniques
- Advanced Optical Sensing Technologies
- Advanced Image Fusion Techniques
- Advanced Neural Network Applications
- Domain Adaptation and Few-Shot Learning
- Advanced Image and Video Retrieval Techniques
- Advanced Electron Microscopy Techniques and Applications
- Video Surveillance and Tracking Methods
- Generative Adversarial Networks and Image Synthesis
- Multimodal Machine Learning Applications
- Computer Graphics and Visualization Techniques
- Visual Attention and Saliency Detection
- Optical Coherence Tomography Applications
- Advanced Fluorescence Microscopy Techniques
- Sparse and Compressive Sensing Techniques
- Image and Video Quality Assessment
- Photoacoustic and Ultrasonic Imaging
- AI in cancer detection
- Digital Media Forensic Detection
University of Science and Technology of China
2016-2025
Nanyang Technological University
2023-2024
Jiangxi Agricultural University
2021-2024
Beijing Institute of Technology
2024
National Science Center
2022-2024
Institute of Art
2022-2024
Xinjiang University
2024
University of Houston
2024
Panzhihua Central Hospital
2024
Central South University
2024
The greatest challenge facing visual object tracking is the simultaneous requirements on robustness and discrimination power. In this paper, we propose a SiamFC-based tracker, named SPM-Tracker, to tackle challenge. basic idea address two in separate matching stages. Robustness strengthened coarse (CM) stage through generalized training while power enhanced fine (FM) distance learning network. stages are connected series as input proposals of FM generated by CM stage. They also parallel...
Time-frequency (T-F) domain masking is a mainstream approach for single-channel speech enhancement. Recently, focuses have been put to phase prediction in addition amplitude prediction. In this paper, we propose phase-and-harmonics-aware deep neural network (DNN), named PHASEN, task. Unlike previous methods which directly use complex ideal ratio mask supervise the DNN learning, design two-stream network, where stream and are dedicated We discover that two streams should communicate with each...
Hyperspectral recovery from a single RGB image has seen great improvement with the development of deep convolutional neural networks (CNNs). In this paper, we propose two advanced CNNs for hyperspectral reconstruction task, collectively called HSCNN+. We first develop residual network named HSCNN-R, which comprises number blocks. The superior performance model comes modern architecture and optimization by removing hand-crafted upsampling in HSCNN. Based on promising results another distinct...
This paper presents a unified deep learning framework to recover hyperspectral images from spectrally undersampled projections. Specifically, we investigate two kinds of representative projections, RGB and compressive sensing (CS) measurements. These measurements are first upsampled in the spectral dimension through simple interpolation or CS reconstruction, proposed method learns an end-to-end mapping large number up-sampled/groundtruth image pairs. The is represented as convolutional...
Existing methods for single image super-resolution (SR) are typically evaluated with synthetic degradation models such as bicubic or Gaussian downsampling. In this paper, we investigate SR from the perspective of camera lenses, named CameraSR, which aims to alleviate intrinsic tradeoff between resolution (R) and field-of-view (V) in realistic imaging systems. Specifically, view R-V a latent model process learn reverse it low- high-resolution pairs. To obtain paired images, propose two novel...
We consider the tracking problem as a special type of object detection problem, which we call instance detection. With proper initialization, detector can be quickly converted into tracker by learning new from single image. find that model-agnostic meta-learning (MAML) offers strategy to initialize satisfies our needs. propose principled three-step approach build high-performance tracker. First, pick any modern trained with gradient descent. Second, conduct offline training (or...
Coded aperture snapshot spectral imaging (CASSI) provides an efficient mechanism for recovering 3D data from a single 2D measurement. However, since the reconstruction problem is severely underdetermined, quality of recovered usually limited. In this paper we propose novel dual-camera design to improve performance CASSI while maintaining its advantage. Specifically, beam splitter placed in front objective lens CASSI, which allows same scene be simultaneously captured by grayscale camera....
Leveraging the compressive sensing (CS) theory, coded aperture snapshot spectral imaging (CASSI) provides an efficient solution to recover 3D hyperspectral data from a 2D measurement. The dual-camera design of CASSI, by adding uncoded panchromatic measurement, enhances reconstruction fidelity while maintaining advantage. In this paper, we propose adaptive nonlocal sparse representation (ANSR) model boost performance (DCCHI). Specifically, CS problem is formulated as cube based make full use...
Liver cancer is a leading cause of deaths worldwide due to its high morbidity and mortality. Histopathological image analysis (HIA) crucial step in the early diagnosis liver routinely performed manually. However, this process time-consuming, error-prone, easily affected by expertise pathologists. Recently, computer-aided methods have been widely applied medical analysis; however, current studies not yet focused on histopathological morphology complex features insufficiency training images...
This paper reviews the NTIRE 2019 challenge on real image denoising with focus proposed methods and their results. The has two tracks for quantitatively evaluating performance in (1) Bayer-pattern raw-RGB (2) standard RGB (sRGB) color spaces. had 216 220 registered participants, respectively. A total of 15 teams, proposing 17 methods, competed final phase challenge. by teams represent current state-of-the-art targeting noisy images.
Contrast enhancement and noise removal are coupled problems for low-light image enhancement. The existing Retinex based methods do not take the coupling relation into consideration, resulting in under or over-smoothing of enhanced images. To address this issue, paper presents a novel progressive framework, which illumination perceived mutually reinforced manner, leading to reduction results. Specifically, two fully pointwise convolutional neural networks devised model statistical...
We propose a Deep Boosting Framework (DBF) for real-world image denoising by integrating the deep learning technique into boosting algorithm. The DBF replaces conventional handcrafted units elaborate convolutional neural networks, which brings notable advantages in terms of both performance and speed. design lightweight Dense Dilated Fusion Network (DDFN) as an embodiment unit, addresses vanishing gradients during training due to cascading networks while promoting efficiency limited...
This paper reviews the first challenge on spectral image reconstruction from RGB images, i.e., recovery of whole-scene hyperspectral (HS) information a 3-channel image. The was divided into 2 tracks: "Clean" track sought HS noiseless images obtained known response function (representing spectrally-calibrated camera) while "Real World" challenged participants to recover cubes JPEG-compressed generated by an unknown function. To facilitate challenge, BGU Hyperspectral Image Database [4]...
Deep learning-based methods especially using convolutional neural network (CNN) and generative adversarial (GAN) have achieved certain success for the task of image inpainting. The previous usually try to generate content in missing areas from scratch. However, these difficulty producing salient structures that appear natural consistent with neighborhood, when area is large. In this paper, we address challenge by introducing edges into GAN-based We split inpainting two steps: first edge...
Event cameras, which output events by detecting spatio- temporal brightness changes, bring a novel paradigm to image sensors with high dynamic range and low latency. Previous works have achieved impressive performances on event-based video reconstruction introducing convolutional neural networks (CNNs). However, intrinsic locality of operations is not capable modeling long-range dependency, crucial many vision tasks. In this paper, we present hybrid CNN- Transformer network for (ET-Net),...
Motion blur is a common photography artifact in dynamic environments that typically comes jointly with the other types of degradation. This paper reviews NTIRE 2021 Challenge on Image Deblurring. In this challenge report, we describe specifics and evaluation results from 2 competition tracks proposed solutions. While both aim to recover high-quality clean image blurry image, different artifacts are involved. track 1, images low resolution while compressed JPEG format. each competition, there...
Deep learning provides a new avenue for light field super-resolution (SR). However, the domain gap caused by drastically different acquisition conditions poses main obstacle in practice. To fill this gap, we propose zero-shot framework SR, which learns mapping to super-resolve reference view with examples extracted solely from input low-resolution itself. Given highly limited training data under setting, however, observe that it is difficult train an end-to-end network successfully. Instead,...
Space-time memory (STM) based video object segmentation (VOS) networks usually keep increasing bank every several frames, which shows excellent performance. However, 1) the hardware cannot withstand ever-increasing requirements as length increases. 2) Storing lots of information inevitably introduces noise, is not conducive to reading most important from bank. In this paper, we propose a Recurrent Dynamic Embedding (RDE) build constant size. Specifically, explicitly generate and update RDE...
Emerging deep learning-based methods have enabled great progress in automatic neuron segmentation from Electron Microscopy (EM) volumes. However, the success of existing is heavily reliant upon a large number annotations that are often expensive and time-consuming to collect due dense distributions complex structures neurons. If required quantity manual for learning cannot be reached, these turn out fragile. To address this issue, article, we propose two-stage, semi-supervised method fully...
Images captured with improper exposures usually bring unsatisfactory visual effects. Previous works mainly focus on either underexposure or overexposure correction, resulting in poor generalization to various exposures. An alternative solution is mix the multiple exposure data for training a single network. However, procedures of correcting and normal are much different from each other, leading large discrepancies network multiple-exposures, thus performance. The key point address this issue...
In this report, we summarize the first NTIRE challenge on light field (LF) image super-resolution (SR), which aims at super-resolving LF images under standard bicubic degradation with a magnification factor of 4. This develops new dataset called NTIRE-2023 for validation and test, provides toolbox BasicLFSR to facilitate model development. Compared single SR, major SR lies in how exploit complementary angular information from plenty views varying disparities. total, 148 participants have...