A5100707809- Image Enhancement Techniques
- Advanced Image Processing Techniques
- Advanced Vision and Imaging
- Digital Imaging for Blood Diseases
- Video Surveillance and Tracking Methods
- AI in cancer detection
- Advanced Measurement and Detection Methods
- Image Processing Techniques and Applications
- Advanced Steganography and Watermarking Techniques
- Advanced Image Fusion Techniques
- Cell Image Analysis Techniques
- Digital Media Forensic Detection
- Face recognition and analysis
- Postharvest Quality and Shelf Life Management
- Remote Sensing and Land Use
- Retinal and Optic Conditions
- Advanced Data Compression Techniques
- Multimodal Machine Learning Applications
- Generative Adversarial Networks and Image Synthesis
- Video Coding and Compression Technologies
- Advanced Image and Video Retrieval Techniques
- Image and Video Stabilization
- Retinal Imaging and Analysis
- Image and Object Detection Techniques
- Ocular Diseases and Behçet’s Syndrome
University of Science and Technology of China
2021-2024
Huawei Technologies (Canada)
2021
Tsinghua University
2020
Xidian University
2017
University of Electronic Science and Technology of China
2011-2015
Zhejiang Gongshang University
2008-2014
National Taichung University of Science and Technology
2013
Xiamen University of Technology
2013
University of Jinan
2011
Peking University
2009-2010
Image demoireing is a multi-faceted image restoration task involving both moire pattern removal and color restoration. In this paper, we raise general degradation model to describe an contaminated by patterns, propose novel multi-scale bandpass convolutional neural network (MBCNN) for single demoireing. For removal, multi-block-size learnable filters (M-LBFs), based on block-wise frequency domain transform, learn the priors of patterns. We also introduce new loss function named Dilated...
Low-light video enhancement (LLVE) is an important yet challenging task with many applications such as photographing and autonomous driving. Unlike single image low-light enhancement, most LLVE methods utilize temporal information from adjacent frames to restore the color remove noise of target frame. However, these algorithms, based on framework multi-frame alignment may produce fusion artifacts when encountering extreme low light or fast motion. In this paper, inspired by latency high...
Ocular images play an essential role in ophthalmological diagnoses. Having imbalanced dataset is inevitable issue automated ocular diseases diagnosis; the scarcity of positive samples always tends to result misdiagnosis severe patients during classification task. Exploring effective computer-aided diagnostic method deal with crucial. In this paper, we develop cost-sensitive deep residual convolutional neural network (CS-ResCNN) classifier diagnose ophthalmic using retro-illumination images....
We investigate GAN inversion problems of using pre-trained GANs to reconstruct real images. Recent methods for such typically employ a VGG perceptual loss measure the difference between While has achieved remarkable success in various computer vision tasks, it may cause unpleasant artifacts and is sensitive changes input scale. This paper delivers an important message that algorithm details are crucial achieving satisfying performance. In particular, we propose two but undervalued design...
Automatic identification of fungi in microscopic fecal images provides important information for evaluating digestive diseases. To date, disease diagnosis is primarily performed by manual techniques. However, the accuracy this approach depends on operator's expertise and subjective factors. The proposed system automatically identifies that contain other cells impurities under complex environments. We segment twice to obtain correct area interest, select ten features, including circle number,...
We propose a novel zero-shot multi-frame image restoration method for removing unwanted obstruction elements (such as rains, snow, and moire patterns) that vary in successive frames. It has three stages: transformer pre-training, restoration, hard patch refinement. Using the pre-trained transformers, our model is able to tell motion difference between true information obstructing elements. For we design model, termed SiamTrans, which constructed by Siamese encoders, decoders. Each temporal...
Text-based person search aims to retrieve the most relevant pedestrian images from an image gallery based on textual descriptions. Most existing methods rely two separate encoders extract and text features, then elaborately design various schemes bridge gap between modalities. However, shallow interaction both modalities in these is still insufficient eliminate modality gap. To address above problem, we propose TransTPS, a transformer-based framework that enables deeper through...
The color variations among different viewpoints in multiview video sequences may deteriorate the visual quality and coding efficiency. Various correction methods have been proposed, however, appearance histogram of corrected target frames are not similar enough to reference details. Focusing on restoring more color, a block-based algorithm is proposed. blocks matched into through spatial prediction, colorization scheme then adopted expand as coarse correction. Finally mixture with global...
Attribute recognition is a crucial but challenging task due to viewpoint changes, illumination variations and appearance diversities, etc. Most of previous work only consider the attribute-level feature embedding, which might perform poorly in complicated heterogeneous conditions. To address this problem, we propose hierarchical embedding (HFE) framework, learns fine-grained by combining attribute ID information. In HFE, maintain inter-class intra-class simultaneously. Not samples with same...
Demoiréing is the task of removing moiré patterns, which are commonly caused by interference between screen and digital cameras. Although research on single image demoiréing has made great progress, video received less attention from community. Video poses a new set challenges. First, most existing restoration algorithms rely multi-resolution pixel-based alignment, can cause damage to details predicted results. Second, these based flow-based loss or relation-based loss, making it difficult...
Camera motion estimation is important for video semantic analysis and classification. can be classified into several classes: camera rotation, translation, zoom, but ordinarily we consider only translation zoom. Traditional approaches parameters features-based, they are dependent on the image quality features' characteristic. not valid in low-density texture images deformed images. In this paper image's algebraic character, introduce a robust approach estimation. Scale parameter extracted...
In order to measure high density print circuit board line widths, an algorithm of image preprocessing is proposed. Methods de-noising are especially researched. This technique can remove noise with adaptive anisotropic diffusion equation filter model based on morphology. Prominent edges be preserved or even enhanced when removed. Experimental results show that the proposed method has strong and preserving edge capability. The accuracy widths measurement improved satisfy real-time detection....