What is Shusheng Puxiang
Shusheng Puxiang is an ultra-high dynamic imaging algorithm developed by Shanghai Artificial Intelligence Laboratory in collaboration with the Chinese University of Hong Kong, Zhejiang University, and other institutions. Its core is Puxiang HDR (UltraFusion HDR). The algorithm combines AIGC technology and HDR technology. Through exposure fusion and generative large models, it can repair image details and generate high-quality images under extreme conditions with exposure differences of up to 9 stops. It effectively solves the motion artifact problem in dynamic scenes and is suitable for fields such as photography, mobile phone imaging, and autonomous driving.
Main functions of Shusheng Puxiang
- Ultra-high dynamic range imaging: Through the combination of AIGC technology and HDR, it can repair image details and generate high-quality images under extreme conditions with exposure differences of up to 9 stops.
- Dynamic scene processing: It effectively solves the motion artifact problem of dynamic objects and is suitable for imaging under complex lighting conditions.
- Multi-exposure fusion: It supports arbitrary exposure input and can fuse images with different exposure levels into a high-quality image with vivid colors and rich details.
Technical principles of Shusheng Puxiang
- Multi-exposure fusion and generative large models: Puxiang HDR is based on AIGC technology and can process images with arbitrary exposure input. Under extreme conditions with exposure differences of up to 9 stops, it can still repair image details and generate high-quality HDR images.
- Dynamic scene processing: Aiming at the motion artifact problem in dynamic scenes, Puxiang HDR adopts an innovative multi-stage processing workflow, effectively solving imaging problems caused by the motion of dynamic objects and ensuring that the generated images are natural and realistic.
- Tone mapping and image restoration: Through the prior knowledge of generative large models, the algorithm adaptively learns tone mapping, overcoming the unnatural tone problem of traditional HDR technology under complex lighting conditions. The guided image restoration design framework can maintain the image fidelity of the generated results and avoid texture change problems.
- Empowered by AIGC technology: AIGC technology provides powerful generative capabilities for HDR imaging, enabling the algorithm to achieve efficient fusion between different exposure levels and significantly improving imaging robustness and color saturation.
Project address of Shusheng Puxiang
- Official homepage：https://ultrafusion.openxlab.org.cn/home
- Project official website：https://openimaginglab.github.io/UltraFusion/
- Online experience Demo：https://huggingface.co/spaces/iimmortall/UltraFusion
Application scenarios of Shusheng Puxiang
- Photography field: It helps photographers repair underexposed or overexposed photos and improve imaging quality without upgrading equipment.
- Mobile photography: It is expected to be integrated into mobile phone cameras to improve mobile phone imaging performance under complex lighting conditions.
- Autonomous driving: In autonomous driving systems, Puxiang HDR can better process images in complex environments and improve the system's perception ability and safety.