Mean Opinion Score (MOS) is the single headline number a subjective test produces: the arithmetic mean of the ratings a panel of viewers gave one clip, almost always on the five-point Absolute Category Rating scale where 5 is excellent and 1 is bad. If N viewers rate a clip, the MOS is the sum of their ratings divided by N. The term came from telephony (ITU-T P.800.1) before video borrowed it. A MOS is only as meaningful as the scale it rides on, and that scale is ordinal, so the categories are ordered but the gaps between them are not guaranteed equal; a 4 is not twice as good as a 2. Crucially, a bare average is not yet a measurement. You compute the MOS only after screening unreliable viewers and you must attach a 95% confidence interval and state the method, scale, and conditions. Every objective metric (PSNR, SSIM, VMAF) is trained against MOS or DMOS.
The handbook · 7 volumes
Fora Soft Video Engineering Handbook
Mean Opinion Score is the human verdict that objective metrics try to predict, but a MOS without enough raters or a confidence interval misleads. We wrote seven volumes covering the whole video stack, and for MOS start with Volume 1: subjective scores, rating scales, VMAF and other metrics trained on MOS data. Written by Nikolay Sapunov, CEO at Fora Soft.
Need MOS numbers that hold up for a codec or vendor decision?
Tell us what you are comparing: we design and run the MOS test, report confidence intervals and give you a clear recommendation. 250+ video products shipped since 2005.
