Aggregating nearest sharp features via hybrid transformers for video deblurring

Wei Shang, Dongwei Ren, Yi Yang, Wangmeng Zuo

Abstract: Highlights•Training a blur-aware detector to distinguish between sharp and blurry frames.•Extracting nearest sharp features from detected frames to guide the restoration.•Using hybrid Transformers to aggregate features from sharp and neighboring frames.•Incorporating an event fusion module to enhance deblurring capabilities.•Performance and efficiency have achieved a good balance on various benchmark datasets.