Therefore, this paper provides a survey of how visual data can be efficiently represented for green multimedia, in particular when the ultimate task is knowledge extraction instead of visual signal ...