MATLAB usage¶

Three reference metrics, ST-MAD, ST-RRED and SpEED-QA, run through MATLAB from the Python harness. They are available only there: the vmaf CLI does not compute them. BRISQUE needs no MATLAB, because libvmaf computes it natively.

Prerequisites¶

Install and activate MATLAB.
Create the optional file compat/python-vmaf/externals.py, which config.py reads, and set MATLAB_PATH in it to your MATLAB binary:
```
MATLAB_PATH = "/path/to/matlab"
```
For example on macOS:
```
MATLAB_PATH = "/Applications/MATLAB_R2017a.app/bin/matlab"
```
To use the MATLAB Runtime instead of a full MATLAB, set MATLAB_RUNTIME_PATH in the same file.

Available algorithms¶

Algorithm	`quality_type`	Notes
ST-MAD [1]	`STMAD`
ST-RRED [2]	`STRRED`
ST-RRED, optimised	`STRREDOpt`	Computationally efficient variant with numerically identical results.
SpEED-QA [3]	`SpEED_Matlab`
BRISQUE [4]	n/a	Native libvmaf feature; see below.

Run the MATLAB algorithms with the run_testing script:

python -m vmaf.script.run_testing <quality_type> <dataset_file>

The dataset file follows the format described in python.md.

BRISQUE¶

BRISQUE runs without MATLAB, as a libvmaf feature extractor:

vmaf --reference ref.yuv --distorted dis.yuv --width 1920 --height 1080 \
     --pixel_format 420 --bitdepth 8 --feature brisque

The trained model ships embedded in the binary. See ../metrics/brisque.md for the options, including an on-disk model such as model/other_models/brisque_live.model.

References¶

[1] P. V. Vu, C. T. Vu, and D. M. Chandler, "A spatiotemporal mostapparent-distortion model for video quality assessment," IEEE Int’l Conf. Image Process., pp. 2505–2508, 2011.

[2] R. Soundararajan and A. C. Bovik, "Video quality assessment by reduced reference spatio-temporal entropic differencing," IEEE Trans. Circ. Syst. Video Technol., vol. 23, no. 4, pp. 684–694, Apr. 2013.

[3] C. G. Bampis, P. Gupta, R. Soundararajan, and A. C. Bovik, "SpEEDQA: Spatial efficient entropic differencing for image and video quality," IEEE Signal Process. Lett., vol. 24, no. 9, pp. 1333–1337, 2017.

[4] A. Mittal, A. K. Moorthy, and A. C. Bovik, "No-reference image quality assessment in the spatial domain," IEEE Trans. Image Process., vol. 21, no. 12, pp. 4695–4708, Dec. 2012.