Objective quality is video quality as computed by an algorithm that reads the pixels or the bitstream and outputs a number, with no humans in the loop. Common metrics are PSNR, which reports raw pixel error in decibels, SSIM, which compares image structure on a 0-to-1 scale, and VMAF, which fuses several measures into a perception-trained 0-to-100 score. Their appeal is that they are fast, deterministic, and repeatable, so they scale to thousands of encodes and power quality gates, regression tests, per-title encoding, and live monitoring. The catch is that an objective score is a model of human opinion, not opinion itself: it is only as trustworthy as its last validation against subjective Mean Opinion Scores, and it can fail on grain, banding, text, dark or high-motion scenes, or footage sharpened to game the number. Objective quality is the scalable proxy for subjective quality, the ground truth it tries to predict.
The handbook · 7 volumes
Fora Soft Video Engineering Handbook
Objective quality is a fast, repeatable stand-in for viewer opinion, and it fails quietly on content it was never validated on. We wrote seven volumes covering the whole video stack, and for objective quality start with Volume 1: PSNR, SSIM and VMAF, subjective scores and where each metric breaks. Written by Nikolay Sapunov, CEO at Fora Soft.
Relying on objective quality scores without checking them against viewers?
Share your content mix and metric results: we validate the objective quality metric on your video and tell you where it can be trusted. 250+ video products shipped since 2005.
