The goal of objective quality assessment (QA) methods is to generate a score for the input visual signal. This score must correlate with human perception of quality.
(i.e. , if we beleive the quality of the image/video under test is fine, the algorithm must generate a higher score for that image/video. )
(The definitons are not precise! more can be found here: https://ocampor.medium.com/advanced-methods-for-iqa-37581ec3c31f ;-) https://en.wikipedia.org/wiki/Structural_similarity )
There are datasets of images/videos along with their human-rated scores. The objective algorithm generates a score for each sample. Then, the correlation between the objective and subjective scores is desired to be high.
This repository doesn't tend to merely automate the testing procedures for different datasets and its ultimate goal is to analyse and visualize the data to gain different insights on methods' performance.