Objective quality measures based on machine learning (ML) require fewer computations and are less affected by inaccuracies in the HVS models. But they may also yield less transparent quality predictions when the ML responses are difficult to interpret. The absence of interpretability may disguise serious vulnerabilities in the design of the objective quality measure.
Barri, A, Dooms, A & Schelkens, P 2014, Reliably combining quality indicators. vol. 1, 2 edn, Video Quality Experts Group, VQEG eLetter. <ftp://vqeg.its.bldrdoc.gov/eLetter/Issues/VQEG_eLetter_vol01_issue2.pdf>
Barri, A., Dooms, A., & Schelkens, P. (2014). Reliably combining quality indicators. (2 ed.) Video Quality Experts Group. ftp://vqeg.its.bldrdoc.gov/eLetter/Issues/VQEG_eLetter_vol01_issue2.pdf
@book{4407a6a1060f41d1943013b240d40a83,
title = "Reliably combining quality indicators",
abstract = "Objective quality measures based on machine learning (ML) require fewer computations and are less affected by inaccuracies in the HVS models. But they may also yield less transparent quality predictions when the ML responses are difficult to interpret. The absence of interpretability may disguise serious vulnerabilities in the design of the objective quality measure.",
author = "Adriaan Barri and Ann Dooms and Peter Schelkens",
year = "2014",
month = dec,
language = "English",
volume = "1",
publisher = "Video Quality Experts Group",
edition = "2",
}