Research

Photography & Imaging (π) Research Group · Nankai University

Research Areas
Low-level Vision

Low-level Vision

We develop algorithms to fundamentally improve image and video quality in degraded conditions. Our work spans a wide spectrum of restoration tasks — from super-resolution and denoising to low-light enhancement, underwater image restoration, and all-in-one degradation removal — with a focus on real-world robustness and perceptual fidelity.

Super-Resolution Denoising Dehazing Deraining Deblurring Face Restoration Low-light Enhancement Underwater Enhancement Color Correction All-in-One Restoration
Representative Publications
CVPR 2026 DNF-SR: Dual-Input and Negative-Aware Feature Fine-Tuning for Real-World Image Super-Resolution
CVPR 2026 Time-Aware One Step Diffusion Network for Real-World Image Super-Resolution
ICCV 2025 DiT4SR: Taming Diffusion Transformer for Real-World Image Super-Resolution
CVPR 2025 Classic Video Denoising in a Machine Learning World: Robust, Fast, and Controllable
View All Papers
Computational Imaging

Computational Imaging

We push the boundaries of camera systems through computational methods. Our research addresses the full imaging pipeline — from RAW sensor data processing and aberration correction to HDR reconstruction and under-display restoration — enabling higher-quality visual capture across diverse hardware platforms and challenging environments.

RAW Denoising & Enhancement HDR Reconstruction Aberration Correction Camera Calibration Flare & Ghost Removal
Representative Publications
NeurIPS 2025 UltraLED: Learning to See Everything in Ultra-High Dynamic Range Scenes
NeurIPS 2024 Lighting Every Darkness in Two Pairs: A Calibration-Free Pipeline for RAW Denoising
CVPR 2023 DNF: Decouple and Feedback Network for Seeing in the Dark
View All Papers
Image Generation and Editing

Image Generation & Editing

We explore generative models for creative visual content production and precise editing. Our research covers identity-preserving video generation, high-resolution video synthesis, semantic image retouching, and efficient diffusion-based editing — bridging the gap between creative intent and photorealistic output.

Video Generation Image Editing Video Inpainting Frame Interpolation Image Colorization Diffusion Models Camera Control
Representative Publications
CVPR 2026 Stand-In: A Lightweight and Plug-and-Play Identity Control for Video Generation
CVPR 2026 YOSE: You Only Select Essential Tokens for Efficient DiT-based Video Object Segmentation
AAAI 2026 VTinker: Guided Flow Upsampling and Texture Mapping for High-Resolution Video Generation
NeurIPS 2025 DIPO: Dual-State Images Controlled Articulated Object Generation Powered by Diverse Data
View All Papers
LLM VLM VLA

LLM / VLM / VLA

We leverage the power of large language and vision-language models to advance image quality assessment, AI-generated content detection, and intelligent visual agents. Our work connects foundation model capabilities with practical imaging applications, enabling smarter, more automated visual workflows.

AI-Generated Content Detection Visual Agents Model Evaluation Benchmark visual-language-action
Representative Publications
AAAI 2026 PerTouch: VLM-Driven Agent for Personalized and Semantic Image Retouching
AAAI 2026 EvalMuse-40K: A Reliable and Fine-Grained Benchmark with Comprehensive Human Annotations for Text-to-Image Generation Model Evaluation
ACM MM 2025 DetectAnyLLM: Towards Generalizable and Robust Detection of Machine-Generated Text
View All Papers
Scene Understanding

Scene Understanding

We tackle high-level visual perception in challenging environments, with a particular focus on underwater scenes. Our research covers object detection, salient object detection, camouflage detection, and instance segmentation — advancing the ability of AI systems to understand complex, real-world visual scenes.

Object Detection Salient Object Detection Camouflage Detection Instance Segmentation Underwater Vision RGB-D Perception
Representative Publications
IEEE TIP 2026 Expose Camouflage in the Water: Underwater Camouflaged Instance Segmentation
IEEE TIP 2025 Heterogeneous Experts and Hierarchical Perception for Underwater Salient Object Detection
View All Papers
3D

3D

We explore 3D scene representation, reconstruction, and understanding. Our research spans 3D Gaussian Splatting for novel-view synthesis and scene reconstruction, depth super-resolution, point cloud robustness, and neural rendering — enabling high-fidelity 3D visual understanding in diverse and challenging scenarios.

3D Gaussian Splatting Neural Rendering Point Cloud Novel View Synthesis RGB-D
Representative Publications
IEEE TIP 2026 3D-UIR: 3D Gaussian for Underwater 3D Scene Reconstruction via Physics-Based Appearance-Medium Decoupling
ICCV 2025 Joint Semantic and Rendering Enhancements in 3D Gaussian Modeling with Anisotropic Local Encoding
IEEE TIP 2025 Improving Point Cloud Robustness Through Perturbation Simulation and Distortion-Tolerant Point Cloud Analysis
View All Papers