Low-Complexity Video Quality Assessment Based on Spatio-Temporal Structure

Low-complexity is as important as prediction accuracy for video quality assessment (VQA) metrics to be practically deployable. In this paper, we develop an effective and efficient full-reference VQA algorithm, called Spatiotemporal Structural-based Video Quality Metric (SSVQM). To be more specific, spatio-temporal structural information is sensitive to both spatial distortions and temporal distortions. We calculate spatio-temporal structure based local quality according to spatio-temporal gradient characteristics and chrominance information. Then, these local quality scores are integrated to yield an overall video quality via a spatio-temporal pooling strategy simulating three most important global temporal effects of the human visual system, i.e. the smooth effect, the asymmetric tracking effect. Experiments on VQA databases LIVE and CSIQ demonstrate that our SSVQM achieves highly competitive prediction accuracy and delivers very low computational complexity.