Sharpness, one of the most effective factors in video quality assessment, usually dominates the first impression of the representation of the compressed video or image signals. In this paper, a new sharpness metric is presented. Without the original video sequence, this metric evaluates the level of sharpness of a compressed video sequence based on the presence of high frequency signals components. Also, an attention module and several human visual factors are included in order to make the measurement results more correlated to human perception. Finally, psychovisual experiments show high correlation between the metric prediction and subjective ranking of video sharpness.
Kai-Chieh Yang, Clark C. Guest, Pankaj K. Das