- Advanced Vision and Imaging
- Advanced Data Compression Techniques
- Video Coding and Compression Technologies
- Image and Signal Denoising Methods
- Advanced Image Processing Techniques
- Generative Adversarial Networks and Image Synthesis
- Digital Filter Design and Implementation
- Sparse and Compressive Sensing Techniques
- Robotic Path Planning Algorithms
- Human Pose and Action Recognition
- Digital Media and Visual Art
- Image and Video Quality Assessment
- Anomaly Detection Techniques and Applications
- Advanced Image and Video Retrieval Techniques
- Cloud Computing and Remote Desktop Technologies
- Multimedia Communication and Technology
- Gait Recognition and Analysis
- Data Management and Algorithms
- Hand Gesture Recognition Systems
- Interactive and Immersive Displays
- Advanced Neural Network Applications
- Video Surveillance and Tracking Methods
- Robotics and Sensor-Based Localization
- Blind Source Separation Techniques
- Image Enhancement Techniques
Chinese Academy of Sciences
2016-2024
Microsoft Research Asia (China)
2010-2024
Aerospace Information Research Institute
2024
Anhui University
2024
Gannan Normal University
2024
Anhui Institute of Information Technology
2020-2023
University of Science and Technology of China
2023
Shanghai Center for Brain Science and Brain-Inspired Technology
2023
Tianjin University
2020
Xi'an University of Technology
2019
<para xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> Nearly all block-based transform schemes for image and video coding developed so far choose the 2-D discrete cosine (DCT) of a square block shape. With almost no exception, this conventional DCT is implemented separately through two 1-D transforms, one along vertical direction another horizontal direction. In paper, we develop new framework in which first may to follow other than or one. The...
Physical objects have inertia, which resists changes in the velocity and motion direction. Inspired by this, we introduce inertia prior that optical flow, reflects object a local temporal window, keeps unchanged adjacent preceding or subsequent frame. We propose flow completion network to align aggregate features from consecutive sequences based on prior. The corrupted flows are completed under supervision of customized losses reconstruction, smoothness, consistent ternary census transform....
The emergence of Kinect facilitates the possibility depth capture in real-time and with low cost by consumers. It also provides powerful tool inspiration for researchers to engage new array technology development. However, quality map captured from is still inadequate many applications due holes, noises artifacts existing within information. In this paper, we present a texture assisted inpainting framework, aiming at obtaining improved relationship between investigated, characteristics are...
Unlike traditional RGB video, Kinect-like depth is characterized by its large variation range and instability. As a result, video compression algorithms cannot be directly applied to with respect coding efficiency. In this paper, we propose lossy framework based on the existing codecs, aiming enhance efficiency while preserving features for further applications. proposed framework, reformed first divisive normalized bilateral filter (DNBL) suppress noises caused disparity normalization, then...
Accuracy and stability of Kinect-like depth data is limited by its generating principle. In order to serve further applications with high quality depth, the preprocessing on essential. this paper, we analyze characteristics examing generation principle propose a spatial-temporal denoising algorithm taking into account special properties. Both intra-frame spatial correlation inter-frame temporal are exploited fill hole suppress noise. Moreover, divisive normalization approach proposed assist...
With the rural revitalization strategy's deepening, North China's mountainous economy needs to shift from traditional modern efficient agriculture. This paper proposes a robust optimization-based optimal crop planting strategy model for region's unique conditions. It first introduces area's topography, climate, and arable land resources, then presents combined linear programming optimization model. Model performance is validated through data preprocessing market analysis, its effectiveness...
Targeting for detecting anomalies of various sizes complicated normal patterns, we propose a Template-guided Hierarchical Feature Restoration method, which introduces two key techniques, bottleneck compression and template-guided compensation, anomaly-free feature restoration. Specially, our framework compresses hierarchical features an image by structure to preserve the most crucial shared among samples. We design compensation restore distorted towards features. Particularly, choose similar...
Transformers have been widely used for video processing owing to the multi-head self attention (MHSA) mechanism. However, MHSA mechanism encounters an intrinsic difficulty inpainting, since features associated with corrupted regions are degraded and incur inaccurate attention. This problem, termed query degradation, may be mitigated by first completing optical flows then using guide attention, which was verified in our previous work – flow-guided transformer (FGT). We further exploit flow...
Nearly all block-based transform schemes for image and video coding developed so far choose the 2-D discrete cosine (DCT) of a square block shape. With almost no exception, this conventional DCT is implemented separately through two 1-D transforms, one along vertical direction another horizontal direction. In paper, we develop new framework in which first may follow other than or one, while second arranged to be one. Compared DCT, resulting directional able provide better performance blocks...
Objects for detection usually have distinct characteristics in different sub-regions and aspect ratios. However, prevalent two-stage object methods, Region-of-Interest (RoI) features are extracted by RoI pooling with little emphasis on these translation-variant feature components. We present selective networks to reform the representations of RoIs exploiting their disparities among Our network produces sub-region attention bank ratio whole image. The RoI-based map selectively pooled from...
Measured load data play a crucial role in the fatigue durability analysis of mechanical structures. However, process signal acquisition, time domain signals are easily contaminated by noise. In this paper, denoising method based on variational mode decomposition (VMD), wavelet threshold (WTD), and singular spectrum (SSA) is proposed. Firstly, simple criterion mutual information entropy (MIE) designed to select proper number for VMD. Detrended fluctuation (DFA) adopted obtain noise level...
A high-peak-power, widely tunable range long-wave infrared optical parametric oscillator (OPO) based on the BaGa 4 Se 7 (BGSe) crystal is demonstrated in this Letter. Pumped by a 1064 nm Nd:YAG laser, high-peak-power of 0.15 MW was achieved at 9.8 µm with pulse width 5.0 ns. At 11.0 µm, high beam quality M 2 x = 4.1 and y 3.3 achieved. By rotating BGSe crystal, broad tuning 6.7–13.9 realized. Furthermore, theoretical analysis conducted to elucidate reasons behind improvement -direction as...
Background. In Traditional Chinese Medicine (TCM), most of the algorithms are used to solve problems syndrome diagnosis that only focus on one syndrome, is, single label learning. However, in clinical practice, patients may simultaneously have more than which has its own symptoms (signs). Methods. We employed a multilabel learning using relevant feature for each (REAL) algorithm construct diagnostic model chronic gastritis (CG) TCM. REAL combines selection methods select significant (signs)...
This paper proposes to use a bipartite graph represent compressive sensing (CS). The evolution of nodes and edges in the graph, which is equivalent decoding process sensing, characterized by set differential equations. One main contributions this that we derive close-form formulation statistics, enable us more accurately analyze performance sensing. Based on formulation, distortion random sampling rate needed code measurements are analyzed briefly. Finally, numerical experiments verify our...
A 3D host–guest Mg-CP exhibits reversible photochromic behavior and emits white light after being post-modified with CuI.
Synthesizing anomaly samples has proven to be an effective strategy for self-supervised 2D industrial detection. However, this approach been rarely explored in multi-modality detection, particularly involving 3D and RGB images. In paper, we propose a novel dual-modality augmentation method synthesis, which is simple capable of mimicking the characteristics defects. Incorporating with our synthesis method, introduce reconstruction-based discriminative detection network, dual-modal...
Compound video compression is crucial for remote control and data assessment. In this paper, we propose a content-aware layered coding scheme as an attempt to efficiently compress the compound video. scheme, analyzed processed progressively at three pyramid levels: block, object layer. Firstly, by block type classification technique access each block's spatial temporal properties. Secondly, natural detected adaptively in frame based on type. Finally, content distributed into different layers...
The pervasive computing environment and wide network bandwidth provide users more opportunities to share screen content among multiple devices. In this article, we introduce a remote display system enable sharing devices with high fidelity responsive interaction. the developed system, frame-level is compressed transmitted client side for sharing, instant control inputs are simultaneously server Even if responds immediately messages updates at frame rate on side, it difficult update low delay...
In this paper, we propose a high frame rate screen video compression scheme aiming at improving the interactive user experience on sharing applications. The proposed is performed as two-layer coding: base layer coding using conventional codec and an enhancement open-loop scheme. For efficient level selection compression, content update of each evaluated through global motion detection. with significant fed to encoder in layer. contrast, little compressed which duplicate indicated by vector...
The Kinect-like depth compression becomes increasingly important due to the growing requirement on Kinect data transmission and storage. Considering temporal inconsistency of introduced by random measurement error, we propose 2D+T prediction algorithm aiming at fully exploiting correlation enhance efficiency. In our prediction, each block is treated as a subsurface, it motion trend detected comparing with reliable 3D reconstruction surface, which integrated accumulated information stored in...
Effective spatiotemporal feature representation is crucial to the video-based action recognition task. Focusing on discriminate learning, we propose Information Fused Temporal Transformation Network (IF-TTN) for top of popular Segment (TSN) framework. In network, Fusion Module (IFM) designed fuse appearance and motion features at multiple ConvNet levels each video snippet, forming a short-term descriptor. With fused as inputs, Networks (TTN) are employed model middle-term temporal...
Transformers have been widely used for video processing owing to the multi-head self attention (MHSA) mechanism. However, MHSA mechanism encounters an intrinsic difficulty inpainting, since features associated with corrupted regions are degraded and incur inaccurate attention. This problem, termed query degradation, may be mitigated by first completing optical flows then using guide attention, which was verified in our previous work - flow-guided transformer (FGT). We further exploit flow...
Porcelain is a precious historical and cultural heritage of China the world as whole,as well treasure inherited from ancient Chinese art. Nevertheless, due to human activities, environmental changes other factors, multitude porcelain relics are undergoing destroy. Therefore, how use modern science technology effectively inherit protect this has become major concern in society. Digital protection attracted board attention with progress 3D scanning modeling printing technology. Double-ear...