Spatio-temporal quality pooling adaptive to distortion distribution and visual attention

In this paper, we propose an adaptive spatio-temporal quality pooling scheme for video quality assessment (VQA) considering the uniformity of distortion distribution and visual attention. A uniformity index is defined according to the distribution of localized distortions in one frame, and is utilized to generate a saliency map from the visual attention model, to incorporate temporal memory effect, and to adaptively pool the localized spatio-temporal quality indices. The proposed scheme outperforms the traditional mean-based pooling scheme and shows competitive performance against well-known VQA algorithms when tested on two publicly available video quality databases.