A decibel (dB) is the logarithmic unit in which PSNR is reported, and understanding it removes most of the mystery from a PSNR figure. PSNR converts the Mean Squared Error into decibels by taking ten times the base-10 logarithm of the squared peak pixel value divided by the error, compressing a huge range of raw error into a friendly two-digit number where higher is better. Because the scale is logarithmic, equal ratios of error become equal additive steps in dB: every halving of the error adds about 3 dB (ten times the log of two is roughly 3.01), every doubling subtracts about 3 dB, so a 6 dB gap is roughly a quarter of the error. That 3-dB-per-doubling rule is the most useful thing to carry. For 8-bit lossy video, PSNR typically lands between 30 and 50 dB, about 48 dB the finite ceiling short of a perfect match. The catch is that the tidy decibel number says nothing about where the error landed perceptually — PSNR's own limitation.
The handbook · 7 volumes
Fora Soft Video Engineering Handbook
Decibels make PSNR easy to misread, because a few dB can mean a large or tiny change depending on where you start. We wrote seven volumes covering the whole video stack, and for reading decibels start with Volume 1: PSNR, SSIM and VMAF, rate-quality curves and BD-PSNR comparisons between H.264, HEVC and AV1. Written by Nikolay Sapunov, CEO at Fora Soft.
Comparing encoders in decibels and unsure what a 1 dB gap means for viewers?
Send us your PSNR results: we translate decibel differences into visible quality and bitrate terms and tell you which gaps matter. 250+ video products shipped since 2005.
