A5045514825- Advanced Image and Video Retrieval Techniques
- Advanced Image Processing Techniques
- Image Retrieval and Classification Techniques
- Image Processing Techniques and Applications
- Advanced Vision and Imaging
- Image and Signal Denoising Methods
- Advanced Image Fusion Techniques
- AI in cancer detection
- Chaos-based Image/Signal Encryption
- Advanced Steganography and Watermarking Techniques
- Advanced Neural Network Applications
- Domain Adaptation and Few-Shot Learning
- Radiomics and Machine Learning in Medical Imaging
- Video Surveillance and Tracking Methods
- Medical Image Segmentation Techniques
- Face and Expression Recognition
- Digital Imaging for Blood Diseases
- Multimodal Machine Learning Applications
- Image Enhancement Techniques
- Face recognition and analysis
- Anomaly Detection Techniques and Applications
- Metaheuristic Optimization Algorithms Research
- Retinal Imaging and Analysis
- Adversarial Robustness in Machine Learning
- Cryptography and Data Security
Guilin University of Electronic Technology
2017-2025
Guilin University
2021-2024
Beijing Satellite Navigation Center
2024
Nanjing University of Aeronautics and Astronautics
2023
Brown University
2023
Guangzhou University
2022
Hong Kong University of Science and Technology
2022
University of Hong Kong
2022
South China University of Technology
2017-2021
East China Normal University
2020
Recently, deep convolutional neural networks (CNNs) have been successfully applied to the single-image super-resolution (SISR) task with great improvement in terms of both peak signal-to-noise ratio (PSNR) and structural similarity (SSIM). However, most existing CNN-based SR models require high computing power, which considerably limits their real-world applications. In addition, methods rarely explore intermediate features that are helpful for final image recovery. To address these issues,...
Chinese image description generation tasks usually have some challenges, such as single-feature extraction, lack of global information, and detailed the content. To address these limitations, we propose a fuzzy attention-based DenseNet-BiLSTM captioning method in this article. In proposed method, first improve densely connected network to extract features at different scales enhance model’s ability capture weak features. At same time, bidirectional LSTM is used decoder use context...
The term "metaverse", a three-dimensional virtual universe similar to the real realm, has always been full of imagination since it was put forward in 1990s. Recently, is possible realize metaverse with continuous emergence and progress various technologies, thus attracted extensive attention again. It may bring lot benefits human society such as reducing discrimination, eliminating individual differences, socializing. However, everything security privacy concerns, which no exception for...
Deep convolutional neural networks (CNNs) have contributed to the significant progress of single-image super-resolution (SISR) field. However, majority existing CNN-based models maintain high performance with massive parameters and exceedingly deeper structures. Moreover, several algorithms essentially underused low-level features, thus causing relatively low performance. In this article, we address these problems by exploring two strategies based on novel local wider residual blocks (LWRBs)...
Collaborative representation is an effective way to design classifiers for many practical applications. In this paper, we propose a novel classifier, called the prior knowledge-based probabilistic collaborative representation-based classifier (PKPCRC), visual recognition. Compared with existing which use strategy, proposed PKPCRC further includes characteristics of training samples each class as knowledge. Four types knowledge are developed from perspectives image distance and capacity. They...
With the popularity of cloud storage services, people are increasingly accustomed to storing images in cloud. However, services raise privacy concerns, e.g., leakage unauthorized third parties and service providers may exploit image detection technologies portrait users without permission. Although concerns can be solved by encrypting before they uploaded cloud, traditional encryption methods significantly affect usability user experience, for example, cannot preview Recently, Marohn <i>et...
This paper proposes a local descriptor called quaternionic ranking binary pattern (QLRBP) for color images. Different from traditional descriptors that are extracted each channel separately or vector representations, QLRBP works on the representation (QR) of image encodes pixel using quaternion. is able to handle all channels directly in domain and include their relations simultaneously. Applying Clifford translation QR image, uses reference quaternion rank QRs two pixels, performs coding...
The tradeoff between efficiency and model size of the convolutional neural network (CNN) is an essential issue for applications CNN-based algorithms to diverse real-world tasks. Although deep learning-based methods have achieved significant improvements in image super-resolution (SR), current techniques mainly contain massive parameters a high computational complexity, limiting their practical applications. In this paper, we present fast lightweight framework, named weighted multi-scale...
Recent advances in single image super-resolution (SISR) demonstrate the power of deep learning for achieving better performance. Because it is costly to recollect training data and retrain model infrared (IR) super-resolution, availability only a few samples restoring IR images presents an important challenge field SISR. To solve this problem, we first propose progressive generative adversarial network (PSRGAN) that includes main path branch path. The depthwise residual block (DWRB) used...
Privacy concerns may be caused after uploading images into image hosting platforms. For traditional encryption, with rich content and meaning are transformed noise-like encrypted without any visual information, which can protect privacy from infringement but at the expense of usability. Recently, thumbnail-preserving encryption (TPE) was proposed to balance usability by preserving thumbnail unchanged encrypting. On other hand, a large number pseudo-random functions high computational...
In this paper, we develop a simple yet powerful framework called quaternion-Michelson descriptor (QMD) to extract local features for color image classification. Unlike traditional descriptors extracted directly from the original (raw) space, QMD is derived Michelson contrast law and quaternionic representation (QR) of images. The stable measurement contents viewpoint human perception, while QR able handle all information holisticly preserve interactions among different channels. way,...
This paper proposes the pairwise linear regression classification (PLRC) for image set retrieval. In PLRC, we first define a new concept of unrelated subspace and introduce two strategies to constitute subspace. order increase information maximizing query set, combination metric classifiers based on constitution Extensive experiments six well-known databases prove that performance PLRC is better than DLRC several state-of-theart different vision recognition tasks: clusterbased face...
In this paper, we explore a new road for format-compatible 3D object encryption by proposing novel mechanism of leveraging 2D image methods. It alleviates the difficulty designing schemes coming from intrinsic intricacy data structure, and implements flexible diverse designs. First, turning complexity into simplicity, vertex values, real numbers with continuous are converted integers ranging 0 to 255. The simplification result is numerical matrix. Second, six prototypes three patterns...
In the field of gland segmentation in histopathology, deep-learning methods have made significant progress. However, most existing not only require a large amount high-quality annotated data but also tend to confuse internal with background. To address this challenge, we propose new semi-supervised method named DCCL-Seg for segmentation, which follows teacher-student framework. Our approach can be divided into steps. First, design contrastive learning module improve ability student model's...