An impairment metric for video temporal fluctuation Measure

Temporal fluctuations are often observed in digitally compressed videos. However, it is difficult to accurately measure these fluctuation intensities with the traditional peak signal-to-noise ratio (PSNR) since the PSNR only provides a generic quality measure. Although specialized metrics have been proposed for temporal fluctuation measurement, e.g., the sum of squared differences (SSD) and the motion compensated SSD (MCSSD), these first difference based algorithms may falsely treat smooth continuous change of pixel values as temporal fluctuations. To overcome this problem, a motion estimated mean scaled absolute second difference (MEMSASD) is proposed here. The performance of the MEMSASD is examined using a number of video sequences with varying degrees of temporal fluctuations, generated by an H.264/AVC compliant codec using standard test video sequences. Compared with the PSNR and the SSD, the behavior of the MCSSD and the proposed metric provide better reflections of temporal fluctuation intensities as perceived by the human visual system (HVS), in terms of the Pearson correlation coefficient. The MEMSASD metric has an advantage over MCSSD in that it avoids misclassification of temporal fluctuations of pixels with smooth continuous change along the temporal axis.