About this project
Description from the program archive
eSSIM presents a solution to the limitations of the widely adopted SSIM metric in accurately assessing human visual perception and performance. This project aims to implement the eSSIM metric as a native video filter in FFmpeg, building upon the existing SSIM filter while incorporating improvements outlined in research papers and existing implementations. Key objectives include transitioning to rectangular windows for local statistics computation, employing integral images for efficiency gains and adopting a Self-Adaptive Scale Transform (SAST).