Saliency based Assessment of Videos from Frame wise Quality Measures

Video quality assessment aims to compute the formal measure of perceived video degradation when video is passed through a video transmission/processing system. Most of the existing video quality measures extend Image Quality Measures by applying them on each frame and later combining the quality values of each frame to get the quality of the entire video. When combining the quality values of frames, a simple average or in very few metrics, weighted average has been traditionally used. In this work, saliency of a frame has been used to compute the weight required for each frame to obtain the quality value of video. The goal of every objective quality metric is to correlate as closely as possible to the perceived quality, and the objective of saliency is parallel to this as the saliency values should match the human perception. Hence we have experimented by using saliency to get the final video quality. The idea is demonstrated by using a number of state of art quality metrics on some of the benchmark datasets.