Per frame metrics/Average metrics

Per frame metrics/Average metric

Metrics are used for measurement video quality. A video stream is a group of frames (pictures, images). Make assessment video quality to count per-frame metrics or average metrics.

Quality comparison. Per-frame metrics.

For getting per-frame metrics we should take two streams. One of them will be used as a reference and another one for comparing in attitude the first one. As a rule, the reference is an original video.

In general, we use some mathematical models to evaluate video quality and measure the similarity between two images. For example, these models are PSNR, SSIM, and VMAF. They are called metrics. Calculate metrics for each frame of a video sequence. Each of the metrics can be determined for different color space components such as Y - luma, U,V - chroma, or for all together - YUV (Y-PSNR, YUV-SSIM).

Average metric.

The average value metric is the mean for all frames in the definite metric and color space, that is, for example, avg U-SSIM, avg YUV-VMAF.

In the company we use the VQ Probe tool. It is a professional instrument for video quality comparison. The tool allows users to compare different codecs, build per-frame metrics and RD-curves, and calculate BD rate.