Quantifying encoding comparison metric uncertainty via bootstrapping

One or more non-transitory computer readable media including instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of:

• computing a first plurality of quality scores associated with a test encoding configuration based on a plurality of bootstrap quality models, wherein each bootstrap quality model is trained based on a different subset of a training database;
• computing a second plurality of quality scores associated with a reference encoding configuration based on the plurality of bootstrap quality models;
• generating a baseline value for an encoding comparison metric based on a baseline quality model trained based on the training database, the test encoding configuration, and the reference encoding configuration;
• and
• generating a distribution of bootstrap values for the encoding comparison metric based on the first plurality of quality scores and the second plurality of quality scores, wherein the distribution quantifies an accuracy of the baseline value for the encoding comparison metric.

A system, comprising:

• one or more memories storing instructions;
• and
• one or more processors that are coupled to the one or more memories and, when executing the instructions, are configured to:
• compute a first plurality of perceptual quality scores associated with a test codec based on a plurality of bootstrap quality models, wherein each bootstrap quality model is trained based on a different subset of a training database;
• compute a second plurality of perceptual quality scores associated with a reference codec based on the plurality of bootstrap quality models;
• and
• generate a distribution of bootstrap values for an encoding comparison metric based on the first plurality of perceptual quality scores and the second plurality of perceptual quality scores, wherein the distribution quantifies an accuracy of a baseline value for the encoding comparison metric generated by a baseline quality model trained based on the training database.