A5054958308- Face and Expression Recognition
- Sparse and Compressive Sensing Techniques
- Visual Attention and Saliency Detection
- Remote-Sensing Image Classification
- Image and Video Quality Assessment
- Advanced Clustering Algorithms Research
- Advanced Image and Video Retrieval Techniques
- Image and Signal Denoising Methods
- Advanced Image Fusion Techniques
- Video Surveillance and Tracking Methods
- Blind Source Separation Techniques
- Image Retrieval and Classification Techniques
- Tensor decomposition and applications
- Alzheimer's disease research and treatments
- Ocular Diseases and Behçet’s Syndrome
- Bioinformatics and Genomic Networks
- Domain Adaptation and Few-Shot Learning
- Advanced Graph Neural Networks
- Dementia and Cognitive Impairment Research
- Advanced Neural Network Applications
- Retinal Diseases and Treatments
- Machine Learning and ELM
- Health, Environment, Cognitive Aging
- Text and Document Classification Technologies
- Complex Network Analysis Techniques
Ocean University of China
2024-2025
Wuhan University of Technology
2025
Qingdao University
2019-2023
University of Electronic Science and Technology of China
2023
Nanjing University of Aeronautics and Astronautics
2023
Qingdao University of Science and Technology
2023
China University of Petroleum, East China
2022
Institute of Software
2022
Yuncheng University
2017
Southern Illinois University Carbondale
2016-2017
Hyperspectral image (HSI) denoising is challenging not only because of the difficulty in preserving both spectral and spatial structures simultaneously, but also due to requirement removing various noises, which are often mixed together. In this paper, we present a nonconvex low rank matrix approximation (NonLRMA) model corresponding HSI method by reformulating problem using regularizer instead traditional nuclear norm, resulting tighter original sparsity-regularised function. NonLRMA aims...
The low-rank tensor representation (LRTR) has become an emerging research direction to boost the multi-view clustering performance. This is because LRTR utilizes not only pairwise relation between data points, but also view of multiple views. However, there one significant challenge: uses nuclear norm as convex approximation provides a biased estimation rank function. To address this limitation, we propose generalized nonconvex (GNLTA) for subspace clustering. Instead correlation, GNLTA...
Graph and subspace clustering methods have become the mainstream of multi-view due to their promising performance. However, (1) since graph learn graphs directly from raw data, when data is distorted by noise outliers, performance may seriously decrease; (2) use a "two-step" strategy representation affinity matrix independently, thus fail explore high correlation. To address these issues, we propose novel method via learning <underline xmlns:mml="http://www.w3.org/1998/Math/MathML"...
Spectral clustering has found extensive use in many areas. Most traditional spectral algorithms work three separate steps: similarity graph construction; continuous labels learning; discretizing the learned by k-means clustering. Such common practice two potential flaws, which may lead to severe information loss and performance degradation. First, predefined might not be optimal for subsequent It is well-accepted that highly affects results. To this end, we propose automatically learn from...
This paper proposes to utilize supervised deep convolutional neural networks take full advantage of the long-term spatial-temporal information in order improve video saliency detection performance. The conventional methods, which use temporally neighbored frames solely, could easily encounter transient failure cases when clues are less-trustworthy for a long period. To tackle aforementioned limitation, we plan identify those beyond-scope with trustworthy first and then align it current...
The current main stream methods formulate their video saliency mainly from two independent venues, i.e., the spatial and temporal branches. As a complementary component, task for branch is to intermittently focus on those regions with salient movements. In this way, even though overall quality heavily dependent its branch, however, performance of still matter. Thus, key factor improve how further boost these branches efficiently. paper, we propose novel spatiotemporal network achieve such...
To solve the saliency detection problem in RGB-D images, depth information plays a critical role distinguishing salient objects or foregrounds from cluttered backgrounds. As complementary component to color information, quality directly dictates subsequent performance. However, due artifacts and limitation of acquisition devices, obtained varies tremendously across different scenarios. Consequently, conventional selective fusion-based methods may result degraded performance cases containing...
Many similarity-based clustering methods work in two separate steps including similarity matrix computation and subsequent spectral clustering. However measurement is challenging because it usually impacted by many factors, e.g., the choice of metric, neighborhood size, scale data, noise outliers. Thus learned often not suitable, let alone optimal, for In addition, nonlinear exists real world data which, however, has been effectively considered most existing methods. To tackle these...
In this article, we introduce a discriminative ridge regression approach to supervised classification. It estimates representation model while accounting for discriminativeness between classes, thereby enabling accurate derivation of categorical information. This new type extends the existing models, such as ridge, lasso, and group by explicitly incorporating As special case, focus on quadratic that admits closed-form analytical solution. The corresponding classifier is called machine (DRM)....
In recent years deep neural networks have been widely applied to visual saliency detection tasks with remarkable performance improvements. As for the salient object in single image, automatically computed convolutional features frequently demonstrate high discriminative power distinguish foregrounds from its non-salient surroundings most cases. Yet, obstinate feature conflicts still persist, which naturally gives rise learning ambiguity, arriving at massive failure detections. To solve such...
Previous video salient object detection (VSOD) approaches have mainly focused on the perspective of network design for achieving performance improvements. However, with recent slowdown in development deep learning techniques, it might become increasingly difficult to anticipate another breakthrough solely via complex networks. Therefore, this paper proposes a universal scheme obtain further 3% improvement all state-of-the-art (SOTA) VSOD models. The major highlight our method is that we...
In this paper, we propose a novel nonconvex approach to robust principal component analysis for HSI denoising, which focuses on simultaneously developing more accurate approximations both rank and column-wise sparsity the low-rank sparse components, respectively. particular, new method adopts log-determinant approximation <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">l</i> <sub xmlns:xlink="http://www.w3.org/1999/xlink">2,log</sub> norm,...
The existing state-of-the-art (SOTA) video salient object detection (VSOD) models have widely followed short-term methodology, which dynamically determines the balance between spatial and temporal saliency fusion by solely considering current consecutive limited frames. However, methodology has one critical limitation, conflicts with real mechanism of our visual system — a typical long-term methodology. As result, failure cases keep showing up in results SOTA models, becomes major technical...
Multi-view subspace clustering (MVSC) has drawn significant attention in recent study. In this paper, we propose a novel approach to MVSC. First, the new method is capable of preserving high-order neighbor information data, which provides essential and complicated underlying relationships data that not straightforwardly preserved by first-order neighbors. Second, design log-based nonconvex approximations both tensor rank sparsity, are effective more accurate than convex approximations. For...
Abstract All‐inorganic perovskite quantum dots (PQDs) have garnered significant attention for optoelectronic applications due to their high photoluminescence yield (PLQY), narrow emission linewidths, and tunable bandgaps. However, inherent instability under environmental conditions susceptibility surface defects limit practical use. In this study, surface‐functionalized mesoporous silica nanospheres (s‐MSNs) are employed as substrates the in situ nucleation growth of CsPbBr 3 PQDs within...