EvTexture

EvTexture is an event-driven texture enhancement method for video super-resolution, designed for researchers and developers working on video enhancement tasks. It leverages high-frequency details of events to better recover texture regions in video super-resolution. The method introduces a new texture enhancement branch and an iterative texture enhancement module to progressively explore high-temporal-resolution event information for texture restoration. This allows for gradual refinement of texture regions across multiple iterations, leading to more accurate and rich high-resolution details. Experimental results show that EvTexture achieves state-of-the-art performance on four datasets, including Vid4, REDS4, Vimeo-90K-T, and CED. For the Vid4 dataset with rich textures, EvTexture can get up to 4.67dB gain compared with recent event-based methods. EvTexture is particularly useful for applications where video quality is crucial, such as in film production, video editing, and live streaming. Its ability to enhance texture regions and reduce errors in restored frames makes it a valuable tool for professionals seeking high-quality video output.