A5063558053- Image and Signal Denoising Methods
- Advanced Image Fusion Techniques
- Photoacoustic and Ultrasonic Imaging
- Advanced Image Processing Techniques
- Image Enhancement Techniques
- Image Processing Techniques and Applications
- Advanced Vision and Imaging
- Optical Coherence Tomography Applications
- Medical Imaging Techniques and Applications
- Anomaly Detection Techniques and Applications
- Computer Graphics and Visualization Techniques
- Optical measurement and interference techniques
- Color Science and Applications
- Advanced X-ray and CT Imaging
- Human Motion and Animation
- Model Reduction and Neural Networks
- Face and Expression Recognition
- Advanced Memory and Neural Computing
- Human Pose and Action Recognition
- Generative Adversarial Networks and Image Synthesis
- Optical Systems and Laser Technology
- Digital Media Forensic Detection
University of Science and Technology of China
2023-2025
Xiangyang Hospital of Traditional Chinese Medicine
2025
Hefei Institutes of Physical Science
2023-2024
Chinese Academy of Sciences
2023-2024
Institute of Intelligent Machines
2023-2024
The goal of pan-sharpening is to produce a high-spatial-resolution multi-spectral (HRMS) image from low-spatial-resolution (LRMS) counterpart by super-resolving the LRMS one under guidance texture-rich panchromatic (PAN) image. Existing research has concentrated on using spatial information generate HRMS images, but neglected investigate frequency domain, which severely restricts performance improvement. In this work, we propose novel approach, named Multi-Scale Dual-Domain Guidance Network...
Pan-sharpening is essentially a panchromatic (PAN)-guided super-resolution process, primarily focused on enhancing multi-spectral image quality. This methodology intricately incorporates the high-frequency derived from texture-rich PAN images into lower-resolution (LRMS) counterparts. However, current spatial domain techniques frequently face challenges in accurately restoring texture details, while frequency methods lack efficient interaction with domains, thus restricting overall model...
The Diffusion Transformer plays a pivotal role in advancing text-to-image and text-to-video generation, owing primarily to its inherent scalability. However, existing controlled diffusion transformer methods incur significant parameter computational overheads suffer from inefficient resource allocation due their failure account for the varying relevance of control information across different layers. To address this, we propose Relevance-Guided Efficient Controllable Generation framework,...
Pan-sharpening involves reconstructing missing high-frequency information in multi-spectral images with low spatial resolution, using a higher-resolution panchromatic image as guidance. Although the inborn connection frequency domain, existing pan-sharpening research has not almost investigated potential solution upon domain. To this end, we propose novel Frequency Adaptive Mixture of Experts (FAME) learning framework for pan-sharpening, which consists three key components: Separation...
Pan-sharpening involves integrating information from lowresolution multi-spectral and high-resolution panchromatic images to generate counterparts. While recent advancements in the state space model, particularly efficient long-range dependency modeling achieved by Mamba, have revolutionized computer vision community, its untapped potential pan-sharpening motivates our exploration. Our contribution, Pan-Mamba, represents a novel pansharpening network that leverages efficiency of Mamba model...
Generating high fidelity human video with specified identities has attracted significant attention in the content generation community. However, existing techniques struggle to strike a balance between training efficiency and identity preservation, either requiring tedious case-by-case finetuning or usually missing details process. In this study, we present ID-Animator, zero-shot human-video approach that can perform personalized given single reference facial image without further training....
Multi-modal image fusion aims to generate a fused by integrating and distinguishing the cross-modality complementary information from multiple source images. While cross-attention mechanism with global spatial interactions appears promising, it only captures second-order interactions, neglecting higher-order in both channel dimensions. This limitation hampers exploitation of synergies between multi-modalities. To bridge this gap, we introduce Synergistic High-order Interaction Paradigm...
Pan-sharpening, a panchromatic image guided low-spatial-resolution multi-spectral super-resolution task, aims to reconstruct the missing high-frequency information of high-resolution counterpart. Although inborn connection with frequency domain, existing pan-sharpening research has almost investigated potential solution upon thus limiting model performance improvement. To this end, we first revisit degradation process in Fourier space, and then devise Pyramid Dual Domain Injection Network...
RAW to sRGB mapping, which aims convert images from smartphones into RGB form equivalent that of Digital Single-Lens Reflex (DSLR) cameras, has become an important area research. However, current methods often ignore the difference between cell phone and DSLR camera images, a goes beyond color matrix extends spatial structure due resolution variations. Recent directly rebuild mapping via shared deep representation, limiting optimal performance. Inspired by Image Signal Processing (ISP)...
RAW to sRGB mapping, which aims convert images from smartphones into RGB form equivalent that of Digital Single-Lens Reflex (DSLR) cameras, has become an important area research. However, current methods often ignore the difference between cell phone and DSLR camera images, a goes beyond color matrix extends spatial structure due resolution variations. Recent directly rebuild mapping via shared deep representation, limiting optimal performance. Inspired by Image Signal Processing (ISP)...
Multi-modal image fusion integrates complementary information from different modalities to produce enhanced and informative images. Although State-Space Models, such as Mamba, are proficient in long-range modeling with linear complexity, most Mamba-based approaches use fixed scanning strategies, which can introduce biased prior information. To mitigate this issue, we propose a novel Bayesian-inspired strategy called Random Shuffle, supplemented by an theoretically-feasible inverse shuffle...
Image restoration aims to reconstruct the latent clear images from their degraded versions. Despite notable achievement, existing methods predominantly focus on handling specific degradation types and thus require specialized models, impeding real-world applications in dynamic scenarios. To address this issue, we propose Large Model Driven Restoration framework (LMDIR), a novel multiple-in-one image paradigm that leverages generic priors large multi-modal language models (MMLMs) pretrained...
Pan-sharpening involves reconstructing missing high-frequency information in multi-spectral images with low spatial resolution, using a higher-resolution panchromatic image as guidance. Although the inborn connection frequency domain, existing pan-sharpening research has not almost investigated potential solution upon domain. To this end, we propose novel Frequency Adaptive Mixture of Experts (FAME) learning framework for pan-sharpening, which consists three key components: Separation...