GLC based multi label image annotation

Annotation is an effective way to make image categorization and retrieval efficient. An effective method for image annotation is to decompose the problem into several independent single-label problems, but this ignores the correlations among different labels. In this paper, an image annotation system is proposed, that automatically annotates images by combining label correlation mining and visual similarity between the images. This method extracts the visually similar images from a collection of multi labelled images by using Content Based Image Retrieval (CBIR). Then the labels of resulted images are label correlated by using GLC (Grading based Label Correlation) algorithm and the correlated labels are assigned to the image. The proposed framework is applied to IAPR -TC image data sets. Experimental results of this approach indicate that the annotation is effective.