Figure 1
From: Supporting visual quality assessment with machine learning
ML-based image/video quality assessment system. The input (distorted) signal DS and, if available, the original signal (OS) are firstrepresented in a low-dimensional feature space and then mapped into the quality space by means of a ML tool. Path (a) includes the two basic modules: 'feature extractor' and 'prediction system'; path (b) augments the framework with a module that is specifically designed to support the prediction system in the task of making a decision about the metric to be applied.