A5056401979- Advanced Image Processing Techniques
- Advanced Vision and Imaging
- Image and Signal Denoising Methods
- Image Enhancement Techniques
- Image Processing Techniques and Applications
- Advanced Data Compression Techniques
- Sparse and Compressive Sensing Techniques
- Advanced Neural Network Applications
- Image Retrieval and Classification Techniques
- Photoacoustic and Ultrasonic Imaging
- Medical Image Segmentation Techniques
- Brain Tumor Detection and Classification
- Advanced Image and Video Retrieval Techniques
- Video Coding and Compression Technologies
- Advanced Image Fusion Techniques
- Advanced Decision-Making Techniques
- Advanced MRI Techniques and Applications
- Image and Object Detection Techniques
- Remote-Sensing Image Classification
- Software System Performance and Reliability
- Robotics and Sensor-Based Localization
- MXene and MAX Phase Materials
- Image Processing and 3D Reconstruction
- Advanced Manufacturing and Logistics Optimization
- Energetic Materials and Combustion
Taiyuan University of Science and Technology
2015-2025
Guangdong University of Finance
2021-2023
Guangdong University Of Finances and Economics
2021
State Key Laboratory of Robotics and Systems
2020
Harbin Institute of Technology
2020
Beijing Jiaotong University
2012-2019
Southern Medical University
2019
University of Chinese Academy of Sciences
2017-2018
Simon Fraser University
2017
Northeast Agricultural University
2016
In autonomous driving, LiDAR sensors are vital for acquiring 3D point clouds, providing reliable geometric information. However, traditional sampling methods of preprocessing often ignore semantic features, leading to detail loss and ground interference in object detection. To address this, we propose a multi-branch two-stage detection framework using Semantic-aware Multi-branch Sampling (SMS) module multi-view consistency constraints. The SMS includes random sampling, Density Equalization...
Multiple description coding (MDC) is able to stably transmit signal in un-reliable and non-prioritized networks, which has been broadly studied for several decades. However, traditional MDC does not well leverage image's context features generate multiple descriptions. In this paper, we propose a novel standard-compliant convolutional neural network-based framework, efficiently leverages information compress the image. First, generator network (MDGN) designed produce appearance-similar yet...
Manual semen evaluation methods are subjective and time-consuming. In this study, a deep learning algorithmic framework was designed to enable non-invasive multidimensional morphological analysis of live sperm in motion, improve current clinical morphology testing methods, significantly contribute the advancement assisted reproductive technologies. We improved FairMOT tracking algorithm by incorporating distance angle same head movement adjacent frames, as well target detection frame IOU...
In recent years, a large number of researchers have begun to embed convolutional neural networks into traditional Compressive Sensing (CS) reconstruction algorithms. They proposed series Deep Unfolding Networks (DUNs) with the characteristics having good interpretability and high-quality reconstruction. However, most DUNs only use inherent CS model, which leads limited performance. addition, simple cannot well remove noises from features. To address above issues, this paper proposes...
The theory of compressed sensing (CS) has been successfully applied to image compression in the past few years, whose traditional iterative reconstruction algorithm is time-consuming. Fortunately, it reported deep learning-based CS algorithms could greatly reduce computational complexity. In this paper, we propose two efficient structures cascaded networks corresponding different sampling methods process. first network a compatibly (CSRNet), which recovers an from its compressively sensed...
Recently, Generative Adversarial Network (GAN) has been found wide applications in style transfer, image-to-image translation and image super-resolution. In this paper, a color-depth conditional GAN is proposed to concurrently resolve the problems of depth super-resolution color 3D videos. Firstly, given low-resolution image, generative network leverage mutual information enhance each other consideration geometry structural dependency same scene. Secondly, three loss functions, including...
Single image super-resolution (SR) aims at reconstructing high-resolution (HR) images from low-resolution (LR) ones. One of the most key issues is to recover finer details LR images. In this paper, considering importance edge prior for SR, we propose a dual-streams driven encoder-decoder network, which combines stream-based network (edge-EDN) and color stream based (color-EDN) reconstruct HR with more details. Instead utilizing two sub-networks learn information contents respectively,...
In 3D video coding and depth-based image rendering, the distortion of compressed depth often leads to wrong warpping. this paper, by generalizing recent work convolutional neural network (CNN)-based up-sampling, we propose a CNN-based artifact removal scheme, where both color images are used enhance accuracy. The proposed CNN has two sub-networks: joint depth-color sub-network sub-network. During feature extraction, gradient is as input image, while extraction. Such an exchange information...
In three‐dimensional (3D) video, a compressed depth map usually has large distortions along‐boundaries, leading to object deformation and artefacts in synthesised views. A so‐called candidate values based boundary filtering (CVBF) with low computational complexity by only some detected unreliable pixels along the boundaries is proposed. Assuming that smooth regions consist of reliable pixels, CVBF selects an appropriate value replace each pixel on both nearest mean surrounding regions....