Objective Perceptual Video Quality Prediction Using Spatial and Temporal Information Differences

In this paper, objective perceptual video quality models are proposed that use spatial and temporal perceptual information differences for predicting video quality as perceived by human observers. Spatial perceptual information characterizes the complexity and temporal perceptual information quantifies the motion contained in a video. As such, differences in the spatial and temporal perceptual information of a reference video (original) and test video (processed) may be used to predict quality of videos that have undergone encoding, transmission, or other processing. In particular, several video quality prediction functions are derived using curve fitting along with training and validation on data from a publicly available annotated database. The obtained functions provide predicted mean opinion scores as a measure of perceptual quality subject to spatial and temporal perceptual information differences. The analysis of the video quality prediction performance of the proposed models shows that differences in spatial and temporal perceptual information can be used for objective video quality prediction.