The field of image processing and generation is rapidly evolving, with a focus on developing innovative methods for image enhancement, restoration, and synthesis. Recent research has explored the use of deep learning techniques, such as diffusion models and transformers, to improve image quality and generate realistic images. Notably, the development of new architectures and training methods has enabled significant advancements in image super-resolution, low-light image enhancement, and image editing. Furthermore, researchers have proposed novel approaches for image fusion, multimodal image processing, and semantic image synthesis, demonstrating the potential for improved performance and efficiency in various applications. Notable papers include: IRDFusion, which proposes a novel feature fusion framework for multispectral object detection, achieving state-of-the-art performance on several datasets. Dark-ISP, which introduces a lightweight and self-adaptive Image Signal Processing plugin for low-light object detection, enabling seamless end-to-end training and superior results in challenging environments. FS-Diff, which presents a semantic guidance and clarity-aware joint image fusion and super-resolution method, demonstrating superior performance in real-world applications.
Advancements in Image Processing and Generation
Sources
An U-Net-Based Deep Neural Network for Cloud Shadow and Sun-Glint Correction of Unmanned Aerial System (UAS) Imagery
IRDFusion: Iterative Relation-Map Difference guided Feature Fusion for Multispectral Object Detection
Dark-ISP: Enhancing RAW Image Processing for Low-Light Object Detection
Texture-aware Intrinsic Image Decomposition with Model- and Learning-based Priors
Composable Score-based Graph Diffusion Model for Multi-Conditional Molecular Generation
FS-Diff: Semantic guidance and clarity-aware simultaneous multimodal image fusion and super-resolution
Chord: Chain of Rendering Decomposition for PBR Material Estimation from Generated Texture Images
Realism Control One-step Diffusion for Real-World Image Super-Resolution
Ordinality of Visible-Thermal Image Intensities for Intrinsic Image Decomposition
InfGen: A Resolution-Agnostic Paradigm for Scalable Image Synthesis
VQT-Light:Lightweight HDR Illumination Map Prediction with Richer Texture.pdf
Adaptive Sampling Scheduler
Exploring Spectral Characteristics for Single Image Reflection Removal
Runge-Kutta Approximation and Decoupled Attention for Rectified Flow Inversion and Semantic Editing
Using KL-Divergence to Focus Frequency Information in Low-Light Image Enhancement
Edge-Aware Normalized Attention for Efficient and Detail-Preserving Single Image Super-Resolution
Dataset Distillation for Super-Resolution without Class Labels and Pre-trained Models
AutoEdit: Automatic Hyperparameter Tuning for Image Editing