Perceptual quality is how good a video looks to a human viewer as the visual system actually perceives it, regardless of whether any original exists. It is the right question for live broadcast, video calls, and user-generated content, where there is no pristine master to measure against. It contrasts with fidelity, which asks how exactly the output matches the source signal. The two usually agree, which is why they get blurred, but they can split apart: a modern AI upscaler can produce a frame that looks better to viewers than the low-resolution source it was built from, giving high perceptual quality but low fidelity, because it invented plausible detail that was never in the original. The catch is that perceptual quality is what truly matters to viewers yet is hard to capture, since the eye weights brightness over color, forgives errors in busy textures, is drawn to faces and motion, and is nearly blind to detail it is not looking at. Metrics like SSIM and VMAF try to model it; PSNR largely does not.
The handbook · 7 volumes
Fora Soft Video Engineering Handbook
Perceptual quality is how the video looks to people, and it can rise even as fidelity falls when AI enhancement invents detail. We wrote seven volumes covering the whole video stack, and for perceptual quality start with Volume 1: perceptual metrics, VMAF and subjective scores, with Volume 6 on AI enhancement and generative video. Written by Nikolay Sapunov, CEO at Fora Soft.
Unsure whether your metrics capture perceptual quality after AI upscaling?
Share before and after clips: we measure perceptual quality and fidelity side by side and recommend which metric should gate your releases. 250+ video products shipped since 2005.
