A5103046319- Image and Video Quality Assessment
- Visual Attention and Saliency Detection
- Error Correcting Code Techniques
- Advanced Wireless Communication Techniques
- Video Coding and Compression Technologies
- Advanced Vision and Imaging
- Radar Systems and Signal Processing
- Cooperative Communication and Network Coding
- Advanced Image Processing Techniques
- Advanced Data Compression Techniques
- PAPR reduction in OFDM
- Target Tracking and Data Fusion in Sensor Networks
- Advanced Image Fusion Techniques
- Image and Signal Denoising Methods
- Video Surveillance and Tracking Methods
- DNA and Biological Computing
- Advanced Image and Video Retrieval Techniques
- Coding theory and cryptography
- Advanced Data Storage Technologies
- Wireless Communication Networks Research
- Image Enhancement Techniques
- Advanced SAR Imaging Techniques
- Algorithms and Data Compression
- Indoor and Outdoor Localization Technologies
- Face Recognition and Perception
Beihang University
2016-2025
Anhui University of Science and Technology
2023
Taishan Medical University
2016-2019
Yuhuangding Hospital
2019
Geospatial Research (United Kingdom)
2018
Shanghai Eighth People Hospital
2009-2017
Wuhan University
2013-2017
University of California, Davis
2012
Jiangxi University of Science and Technology
2012
China Astronaut Research and Training Center
2012
Glaucoma is one of the leading causes irreversible vision loss. Many approaches have recently been proposed for automatic glaucoma detection based on fundus images. However, none existing can efficiently remove high redundancy in images detection, which may reduce reliability and accuracy detection. To avoid this disadvantage, paper proposes an attention-based convolutional neural network (CNN) called AG-CNN. Specifically, we first establish a large-scale (LAG) database, includes 11 760...
A deep learning system (DLS) that could automatically detect glaucomatous optic neuropathy (GON) with high sensitivity and specificity expedite screening for GON.To establish a DLS detection of GON using retinal fundus images glaucoma diagnosis convoluted neural networks (GD-CNN) has the ability to be generalized across populations.In this cross-sectional study, classification was developed automated obtained from Chinese Glaucoma Study Alliance, Handan Eye Study, online databases. The...
The past few years have witnessed great success in applying deep learning to enhance the quality of compressed image/video. existing approaches mainly focus on enhancing a single frame, ignoring similarity between consecutive frames. In this paper, we investigate that heavy fluctuation exists across video frames, and thus low frames can be enhanced using neighboring high seen as Multi-Frame Quality Enhancement (MFQE). Accordingly, paper proposes an MFQE approach for video, first attempt...
The past few years have witnessed great success in applying deep learning to enhance the quality of compressed image/video. existing approaches mainly focus on enhancing a single frame, not considering similarity between consecutive frames. Since heavy fluctuation exists across video frames as investigated this paper, frame can be utilized for enhancement low-quality given their neighboring high-quality This task is Multi-Frame Quality Enhancement (MFQE). Accordingly, paper proposes an MFQE...
Panoramic video provides immersive and interactive experience by enabling humans to control the field of view (FoV) through head movement (HM). Thus, HM plays a key role in modeling human attention on panoramic video. This paper establishes database collecting subjects' sequences. From this database, we find that data are highly consistent across subjects. Furthermore, deep reinforcement learning (DRL) can be applied predict positions, via maximizing reward imitating scanpaths agent's...
In this letter, we propose a road structure refined convolutional neural network (RSRCNN) approach for extraction in aerial images. order to obtain structured output of extraction, both deconvolutional and fusion layers are designed the architecture RSRCNN. For training RSRCNN, new loss function is proposed incorporate geometric information cross-entropy loss, thus called road-structure-based function. Experimental results demonstrate that trained RSRCNN model able advance state-of-the-art...
The latest High Efficiency Video Coding (HEVC) standard has been increasingly applied to generate video streams over the Internet. However, HEVC compressed videos may incur severe quality degradation, particularly at low bit-rates. Thus, it is necessary enhance visual of decoder side. To this end, paper proposes a Quality Enhancement Convolutional Neural Network (QE-CNN) method that does not require any modification encoder achieve enhancement for HEVC. In particular, our QE-CNN learns...
High Efficiency Video Coding (HEVC) significantly reduces bit-rates over the proceeding H.264 standard but at expense of extremely high encoding complexity. In HEVC, quad-tree partition coding unit (CU) consumes a large proportion HEVC complexity, due to bruteforce search for rate-distortion optimization (RDO). Therefore, this paper proposes deep learning approach predict CU reducing complexity both intra- and inter-modes, which is based on convolutional neural network (CNN) long- short-term...
An extensive study on the in-loop filter has been proposed for a high efficiency video coding (HEVC) standard to reduce compression artifacts, thus improving efficiency. However, in existing approaches, is always applied each single frame, without exploiting content correlation among multiple frames. In this paper, we propose multi-frame (MIF) HEVC, which enhances visual quality of encoded frame by leveraging its adjacent Specifically, first construct large-scale database containing frames...
In contrast with traditional video, omnidirectional video enables spherical viewing direction support for head-mounted displays, providing an interactive and immersive experience. Unfortunately, to the best of our knowledge, there are few visual quality assessment (VQA) methods, either subjective or objective, coding. This paper proposes both objective methods assessing loss in encoding video. Specifically, we first present a new database, which includes data from several subjects watching...
The latest High Efficiency Video Coding (HEVC) has been increasingly used to generate video streams over Internet. However, the decoded HEVC may incur severe quality degradation, especially at low bit-rates. Thus, it is necessary enhance visual of videos decoder side. To this end, we propose in paper a Decoder-side Scalable Convolutional Neural Network (DS-CNN) approach achieve enhancement for HEVC, which does not require any modification encoder. In particular, our DS-CNN learns model...
For High Efficiency Video Coding (HEVC), the R– <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\lambda $ </tex-math></inline-formula> scheme is latest rate control (RC) scheme, which investigates relationships among allocated bits, slope of rate-distortion (R-D) curve , and quantization parameter. However, we argue that bit allocation in existing not optimal. In this paper, therefore propose an optimal...
Omnidirectional video enables spherical stimuli with the $360 \times 180^ \circ$ viewing range. Meanwhile, only viewport region of omnidirectional can be seen by observer through head movement (HM), and an even smaller within clearly perceived eye (EM). Thus, subjective quality may correlated HM EM human behavior. To fill in gap between behavior, this paper proposes a large-scale visual assessment (VQA) dataset video, called VQA-OV, which collects 60 reference sequences 540 impaired...
Given the outbreak of COVID-19 pandemic and shortage medical resource, extensive deep learning models have been proposed for automatic diagnosis, based on 3D computed tomography (CT) scans. However, existing independently process lesion segmentation disease classification, ignoring inherent correlation between these two tasks. In this paper, we propose a joint model classification diagnosing COVID-19, called DeepSC-COVID, as first attempt in direction. Specifically, establish large-scale CT...
In this paper, we propose a region-of-interest (ROI) based HEVC coding approach for conversational videos, with novel hierarchical perception model of face (HP model), to improve the perceived visual quality state-of-the-art standard. contrast previous ROI-based video approaches, HP allows unequal importance facial features (e.g., eyes and mouth) within region, by generating pixel-wise weight map. Benefitting from such model, adaptive tree unit (CTU) partition structure is developed...
Saliency detection has been widely studied to predict human fixations, with various applications in computer vision and image processing. For saliency detection, we argue this paper that the state-of-the-art High Efficiency Video Coding (HEVC) standard can be used generate useful features compressed domain. Therefore, proposes learn video model, regard HEVC features. First, establish an eye tracking database for which downloaded from https://github.com/remega/video_database. Through...
High efficiency video coding (HEVC) is the latest standard, and it has best performance among all existing standards. HEVC main still picture profile (HEVC-MSP) also achieves top in image compr-ession. In this paper, we propose a closed-form bit allocation approach to optimize saliency-guided PSNR (viewed as perceptual distortion) such that of HEVC-based compression can be significantly improved from subjective perspective. Specifically, formulation established minimize distortion with...
Diabetic retinopathy (DR) is a leading cause of permanent blindness among the working-age people. Automatic DR grading can help ophthalmologists make timely treatment for patients. However, existing methods are usually trained with high resolution (HR) fundus images, such that performance decreases lot given low (LR) which common in clinic. In this paper, we mainly focus on LR images. According to our analysis task, find that: 1) image super-resolution (ISR) boost both and lesion...
In order to achieve high spectral efficiency and low access delay, this paper introduces cyclical non-orthogonal multiple (NOMA) into unmanned aerial vehicle (UAV)-enabled wireless network. It allows the UAV communicate with ground users in same time–frequency resources, cyclically. The minimum throughput over all is maximized by jointly optimizing multiuser communication scheduling NOMA trajectory. turns out that maximization of a mixed integer non-linear non-convex optimization problem....
Saliency prediction in traditional images and videos has drawn extensive research interests recent years. Few works have been proposed for saliency over 360° videos. They focus on directly predicting fixations the whole panorama. When viewing videos, a person can only observe content her viewport, which means that fraction of scene be seen at any given time. In this paper, we study human attention viewport propose novel visual model, dubbed saliency, to predict Two contributions are...