FURI | Summer 2025

Event-Based Deep Learning Model for Advanced Image and Video Deblurring

Blurry visuals are a significant challenge in capturing clear images and videos during fast motion or in poor lighting conditions, affecting crucial applications like robotics, self-driving vehicles, and surveillance systems. Current methods often fail to rapidly or accurately address this blur. This research proposes an innovative solution combining event-based cameras, which capture visual information through motion-triggered pixel intensity changes, with advanced deep learning methods that integrate hybrid convolutional and transformer-based neural networks. By leveraging both image and event-motion data, this approach aims to achieve superior deblurring performance, surpassing existing methods in speed and clarity. Successful implementation will improve visual system reliability across multiple industries, contributing to safer and more effective real-world technology deployment.