A5081607584- Advanced Image Processing Techniques
- Image Enhancement Techniques
- Image and Signal Denoising Methods
- Advanced Vision and Imaging
- Advanced Image Fusion Techniques
- Image Processing Techniques and Applications
- Advanced Neural Network Applications
- Anomaly Detection Techniques and Applications
- Video Surveillance and Tracking Methods
- Domain Adaptation and Few-Shot Learning
- Medical Image Segmentation Techniques
- Human Pose and Action Recognition
- Image and Video Quality Assessment
- Multimodal Machine Learning Applications
- COVID-19 diagnosis using AI
- Radiomics and Machine Learning in Medical Imaging
- AI in cancer detection
- Machine Learning and ELM
- Medical Imaging Techniques and Applications
- Advanced Image and Video Retrieval Techniques
- CCD and CMOS Imaging Sensors
- Photoacoustic and Ultrasonic Imaging
- Advanced Optical Sensing Technologies
- Infrared Target Detection Methodologies
- Network Security and Intrusion Detection
Northwestern Polytechnical University
2014-2025
University of Würzburg
2022-2023
The University of Adelaide
2018-2022
University of Southern Denmark
2022
Huawei Technologies (Sweden)
2022
Australian Centre for Robotic Vision
2021-2022
Beijing Anzhen Hospital
2020
Taiyuan University of Science and Technology
2013-2014
Blind image quality assessment (BIQA) for authentically distorted images has always been a challenging problem, since captured in the wild include varies contents and diverse types of distortions. The vast majority prior BIQA methods focus on how to predict synthetic quality, but fail when applied real-world images. To deal with challenge, we propose self-adaptive hyper network architecture blind assess wild. We separate IQA procedure into three stages including content understanding,...
Ghosting artifacts caused by moving objects or misalignments is a key challenge in high dynamic range (HDR) imaging for scenes. Previous methods first register the input low (LDR) images using optical flow before merging them, which are error-prone and cause ghosts results. A very recent work tries to bypass flows via deep network with skip-connections, however, still suffers from ghosting severe movement. To avoid source, we propose novel attention-guided end-to-end neural (AHDRNet) produce...
One of the most challenging problems in reconstructing a high dynamic range (HDR) image from multiple low (LDR) inputs is ghosting artifacts caused by object motion across different inputs. When slight, existing methods can well suppress through aligning LDR based on optical flow or detecting anomalies among them. However, they often fail to produce satisfactory results practice, since real be very large. In this study, we present novel deep framework, termed NHDRRnet, which adopts an...
The recent contrastive language-image pre-training (CLIP) model has shown great success in a wide range of image-level tasks, revealing remarkable ability for learning powerful visual representations with rich semantics. An open and worthwhile problem is efficiently adapting such strong to the video domain designing robust anomaly detector. In this work, we propose VadCLIP, new paradigm weakly supervised detection (WSVAD) by leveraging frozen CLIP directly without any fine-tuning process....
Traditional image quality assessment (IQA) methods do not perform robustly due to the shallow hand-designed features. It has been demonstrated that deep neural network can learn more effective features than ever. In this paper, we describe a new predict accurately without relying on reference image. To feature representations for non-reference IQA, propose two-stream convolution includes two subcomponents and gradient The motivation design is using scheme capture different-level information...
Liver vessel segmentation is fast becoming a key instrument in the diagnosis and surgical planning of liver diseases. In clinical practice, vessels are normally manual annotated by clinicians on each slice CT images, which extremely laborious. Several deep learning methods exist for segmentation, however, promoting performance remains major challenge due to large variations complex structure vessels. Previous mainly using existing UNet architecture, but not all features encoder useful some...
A novel coronavirus disease 2019 (COVID-19) was detected and has spread rapidly across various countries around the world since end of year 2019, Computed Tomography (CT) images have been used as a crucial alternative to time-consuming RT-PCR test. However, pure manual segmentation CT faces serious challenge with increase suspected cases, resulting in urgent requirements for accurate automatic COVID-19 infections. Unfortunately, imaging characteristics infection are diverse similar...
A novel coronavirus disease 2019 (COVID-19) was detected and has spread rapidly across various countries around the world since end of year 2019. Computed Tomography (CT) images have been used as a crucial alternative to time-consuming RT-PCR test. However, pure manual segmentation CT faces serious challenge with increase suspected cases, resulting in urgent requirements for accurate automatic COVID-19 infections. Unfortunately, imaging characteristics infection are diverse similar...
Generating a high dynamic range (HDR) image from set of sequential exposures is challenging task for scenes. The most common approaches are aligning the input images to reference before merging them into an HDR image, but artifacts often appear in cases large scene motion. state-of-the-art method using deep learning can solve this problem effectively. In paper, we propose novel convolutional neural network generate HDR, which attempts produce more vivid images. key idea our coarse-to-fine...
High Dynamic Range (HDR) images can be recovered from several Low (LDR) by existing Deep Neural Networks (DNNs) techniques. Despite the remarkable progress, DNN-based methods still generate ghosting artifacts when LDR have saturation and large motion, which hinders potential applications in real-world scenarios. To address this challenge, we formulate HDR deghosting problem as an image generation that leverages features diffusion model's condition, consisting of feature condition generator...
Automatic segmentation of liver tumors is crucial to assist radiologists in clinical diagnosis. While various deep learningbased algorithms have been proposed, such as U-Net and its variants, the inability explicitly model long-range dependencies CNN limits extraction complex tumor features. Some researchers applied Transformer-based 3D networks analyze medical images. However, previous methods focus on modeling local information (eg. edge) or global morphology) with fixed network weights....
Diffusion models have gained significant popularity for image-to-image translation tasks. Previous efforts applying diffusion to image super-resolution demonstrated that iteratively refining pure Gaussian noise using a U-Net architecture trained on denoising at various levels can yield satisfactory high-resolution images from low-resolution inputs. However, this iterative refinement process comes with the drawback of low inference speed, which strongly limits its applications. To speed up...
This paper reviews the challenge on constrained high dynamic range (HDR) imaging that was part of New Trends in Image Restoration and Enhancement (NTIRE) workshop, held conjunction with CVPR 2022. manuscript focuses competition set-up, datasets, proposed methods their results. The aims at estimating an HDR image from multiple respective low (LDR) observations, which might suffer under-or over-exposed regions different sources noise. is composed two tracks emphasis fidelity complexity...
Continual Learning (CL) methods aim to enable machine learning models learn new tasks without catastrophic forgetting of those that have been previously mastered. Existing CL approaches often keep a buffer previously-seen samples, perform knowledge distillation, or use regularization techniques towards this goal. Despite their performance, they still suffer from interference across which leads forgetting. To ameliorate problem, we propose only activate and select sparse neurons for current...
This paper introduces a novel benchmark for efficient up-scaling as part of the NTIRE 2023 Real-Time Image Super-Resolution (RTSR) Challenge, which aimed to upscale images from 720p and 1080p resolution native 4K (×2 ×3 factors) in real-time on commercial GPUs. For this, we use new test set containing diverse ranging digital art gaming photography. We assessed methods devised SR by measuring their runtime, parameters, FLOPs, while ensuring minimum PSNR fidelity over Bicubic interpolation....
Mapping Low Dynamic Range (LDR) images with different exposures to High (HDR) remains nontrivial and challenging on dynamic scenes due ghosting caused by object motion or camera jitting. With the success of Deep Neural Networks (DNNs), several DNNs-based methods have been proposed alleviate ghosting, they cannot generate approving results when saturation occur. To visually pleasing HDR in various cases, we propose a hybrid deghosting network, called HyHDRNet, learn complicated relationship...