In this article, a general framework of image annotation is proposed by involving salient object detection (SOD), feature extraction, feature selection, and multi-label classification. For SOD, Augmented-Gradient Vector ...