Motion estimated temporal consistency metrics for objective video quality assessment

Abstract

Video quality evaluation is an important component in improving the user experience in many multimedia applications. Therefore, there is significant interest in generating objective video quality metrics that can accurately estimate the perceived quality of image sequences. Many existing video quality metrics, however, do not explicitly account for temporal artifacts that occur in video compression. This work proposes a set of temporal consistency metrics to help measure the extent of temporal distortion artifacts in video sequences. The proposed metrics perform motion estimation in the input sequences, and measure the temporal consistency in the motion compensated images of the sequence. Global and local weighting parameters are used to ensure that only the perceivable distortions will be measured by the metrics. The temporal consistency metrics are then combined with existing spatial distortion metrics. Initial experimental results show that the combined spatio-temporal distortion metric provides improved correlation with subjective measurements compared to spatial-only metrics.