A Distortion-Weighing Spatiotemporal Visual Attention Model for Video Analysis

For the purpose of extracting the attention regions from the distorted videos, a distortions-weighing spatiotemporal visual attention model is proposed. On the impact of the spatial and the temporal saliency maps, the visual attention regions are acquired directed in a bottom-up manner. Meanwhile, the blockiness saliency map is detected according to the intensity gradient features. An attention selection is applied to identify one of the visual attention regions with more relative serious blockiness as the Focus of Attention (FOA), which is directed in a top-down manner. Experimental results show that the proposed model can not only accurately analyze the spatiotemporal saliency based on the intensity, texture, and motion features, but also able to estimate the blockiness of the distortions in comparing with Walther´s and You´s models.