A Survey and Analysis on Image Annotation

The rapid growth of archives of available visual content, such as photo or video sharing websites, has created a need for indexing techniques and multimedia information search, and more precisely images. Image annotation allows indexing and searching large collections of images in an easy and fast way. Annotation of images by keywords is a way to associate “semantics” with an image. It consists of assigning each image, a keyword or a set of keywords, intended to describe the semantic content of the image. In this paper, a detailed study of image annotation methods are devoted for distinguishing between manual, semi-automatic and automatic annotation. The evaluation measures of annotation systems and the description of utilized databases are also devoted in this survey. This paper conclude with a synthesis of state-of-the-art methods to motivate choices. This synthesis allowed us to argue choice for probabilistic graphic models as a method of annotation and the importance of integrating user feedback and a semantic hierarchy into image annotation model.