Making Video Quality Assessment Models Sensitive to Frame Rate Distortions

We consider the problem of capturing distortions arising from changes in frame rate as part Video Quality Assessment (VQA). Variable (VFR) videos have become much more common, and streamed commonly range 30 frames per second (fps) up to 120 fps. VFR-VQA offers unique challenges terms distortion types well making non-uniform comparisons reference distorted having different rates. The majority current VQA models require compared be same rate, but are unable adequately account for artifacts. recently proposed Generalized Entropic Difference (GREED) model succeeds at this task, using natural video statistics entropic differences temporal band-pass coefficients, delivering superior performance on predicting quality distortions. Here we propose a simple fusion framework, whereby features GREED combined with existing models, towards improving sensitivity find through extensive experiments that feature significantly boosts both HFR/VFR datasets fixed (FFR) databases. Our results suggest employing efficient representations can result robust accurate when variations occur.