The just-noticeable difference (JND) is the smallest change in quality that a typical viewer can actually perceive, and it is useful as a natural unit for quality differences. Its value lies in turning a raw metric gap into something meaningful: a one-point change on a metric scale may sit well below a JND and be invisible, so treating it as a real improvement is reading noise as signal. The concept echoes the confidence-interval idea in metric reading, where two encodes whose score ranges overlap cannot be told apart. JND thinking complements validation statistics like PCC and SROCC by reminding you that a difference must clear a perceptual threshold before it matters to a person, not just clear a numerical threshold. It is also why small score gaps between encodes should be checked against an error bar or, when stakes are high, a subjective test, before any winner is declared. A JND anchors the question of whether a measured difference is one a human would ever notice.
The handbook · 7 volumes
Fora Soft Video Engineering Handbook
The just-noticeable difference tells you whether a metric gap is visible at all, which stops teams from chasing changes nobody can see. We wrote seven volumes covering the whole video stack, and for JND start with Volume 1: human vision, perceptual quality metrics, VMAF score differences and rate-quality trade-offs. Written by Nikolay Sapunov, CEO at Fora Soft.
Chasing a 2-point VMAF gain without knowing if it is a just-noticeable difference?
Share your encodes and metric results: we tell you which differences cross one JND and where you can cut bitrate with no visible loss. 250+ video products shipped since 2005.
