EvTexture++: Event-Driven Texture Enhancement for Video Super-Resolution
Researchers have developed EvTexture++, a novel framework for video super-resolution that utilizes event-based vision data to enhance texture recovery. This approach shifts the focus from motion estimation to texture enhancement, leveraging high-frequency spatiotemporal details from events. EvTexture++ includes a dedicated texture enhancement branch and an iterative module for progressive texture restoration, as well as a temporal texture alignment module to improve inter-frame consistency. The framework is designed to be plug-and-play, enhancing existing video super-resolution models and demonstrating state-of-the-art performance with significant PSNR gains on texture-rich datasets. AI
IMPACT Enhances video super-resolution capabilities by leveraging event-based vision for improved texture detail and temporal consistency.