libvmaf

Calculate the VMAF (Video Multi-Method Assessment Fusion) score for a reference/distorted pair of input videos.

The first input is the distorted video, and the second input is the reference video.

The obtained VMAF score is printed through the logging system.

It requires Netflix’s vmaf library (libvmaf) as a pre-requisite. After installing the library it can be enabled using: ./configure --enable-libvmaf.

The filter has following options:

model

A ‘|‘ delimited list of vmaf models. Each model can be configured with a number of parameters. Default value: "version=vmaf_v0.6.1"

feature

A ‘|‘ delimited list of features. Each feature can be configured with a number of parameters.

log_path

Set the file path to be used to store log files.

log_fmt

Set the format of the log file (xml, json, csv, or sub).

pool

Set the pool method to be used for computing vmaf. Options are min, harmonic_mean or mean (default).

n_threads

Set number of threads to be used when initializing libvmaf. Default value: 0, no threads.

n_subsample

Set frame subsampling interval to be used.

This filter also supports the framesync options.

Examples

• In the examples below, a distorted video distorted.mpg is compared with a reference file reference.mpg.
• Basic usage:

  ffmpeg -i distorted.mpg -i reference.mpg -lavfi libvmaf=log_path=output.xml -f null -
  

• Example with multiple models:

  ffmpeg -i distorted.mpg -i reference.mpg -lavfi libvmaf='model=version=vmaf_v0.6.1\\:name=vmaf|version=vmaf_v0.6.1neg\\:name=vmaf_neg' -f null -
  

• Example with multiple additional features:

  ffmpeg -i distorted.mpg -i reference.mpg -lavfi libvmaf='feature=name=psnr|name=ciede' -f null -
  

• Example with options and different containers:

  ffmpeg -i distorted.mpg -i reference.mkv -lavfi "[0:v]settb=AVTB,setpts=PTS-STARTPTS[main];[1:v]settb=AVTB,setpts=PTS-STARTPTS[ref];[main][ref]libvmaf=log_fmt=json:log_path=output.json" -f null -