PSNR — Peak Signal-to-Noise Ratio — is the oldest and simplest video quality metric. It's a single number, measured in decibels (dB), that compares an encoded video frame to its original source and reports how different they are pixel by pixel. Higher is better; identical videos score infinity. The metric has been around since the 1970s, runs almost instantly even on huge files, and is built into every video tool from FFmpeg to professional broadcast monitors.
What PSNR is good for: spotting catastrophic encoding errors, comparing two encodes of the same source, sanity-checking that a transcode worked. As a rough rule of thumb, PSNR above 40 dB is excellent quality, 30–40 dB is acceptable, below 30 dB is visibly degraded. The numbers are reliable for big quality differences and convenient because every encoding tool reports them.
What PSNR is bad for: predicting what humans will actually perceive. Because it treats every pixel as equally important and just measures raw mathematical error, PSNR can rate two videos identically even when one looks crisp and the other looks like mush — for example, a slight blur and a small added noise pattern can produce the same PSNR but completely different perceived quality. Modern video teams use PSNR as a quick first check but lean on ssim and especially vmaf for decisions about real-world quality. The 2026 standard practice is to report all three side by side.
