The influences of task and scene content on a viewer's subjective opinion of compressed video quality are investigated. A group of test subjects are presented with a number of compressed video clips, with or without an initial "task" instruction. Subjective quality ratings and selected eye movement tracking results are recorded. The results indicate that subjective quality is strongly influenced by the presence or absence of distortion in foreground human figures in a video scene and is also influenced by the presence or absence of an initial task. The implications of these results for subjective quality testing and for the design of video compression systems are discussed.
|Number of pages
|Published - 2004