Measurement of scene-dependent quality variations in digitally coded television pictures

Experiments are reported in which the ITU-R Double Stimulus Continuous Quality Scale (DSCQS) method was used to assess scene-dependent, time-varying television picture coding distortions introduced by MPEG-2 coding, at bit rates ranging from 1 Mbit/s to 7.5 Mbit/s. Test sequence lengths of both 10 s and 30 s were employed, the latter in an attempt to present to subjects a more representative range of quality variations, which were found to be large. In forming quality judgements of 30 s sequences, subjects gave significantly more weight to impairments occurring towards the end of the sequence; for identical impairments in the first and in the last 10 s of such a sequence, the difference in rating was typically about half a scale point on a 5-point scale. This recency effect is probably linked to the short-term nature of human working memory