An Instance-based Method for Automatic Image Semantics Annotation

In content-based image retrieval,annotations of image can not only reduce the gap between high-grade semantics and low-grade visual content,but is also convenient for retrieval.As we all know,manual-annotation is time-consuming,strength-consuming,and the annotation results may be subjectively different,while the automatic image annotation can transform the visual features into annotations.This benefits users a lot.In this paper,we propose an instance-based method for automatic image annotation.The advantage of this method is that it can determine the image annotation according to previous annotation experiences,which overcomes the shortcomings of manual-annotation.Moreover,the method can be extended to semi-automatic annotation,which provides users with a simple and convenient interface for annotation.