Automatic Image Annotation and Retrieval

Abstract

In this paper, an automatic image annotation and retrieval model is developed base on the intensity invariant approach. The given uncaptioned image is divided into background and foreground images and segmented into regions, which are classified into region types using a variety of features. Firstly, preprocessing stages such as gray-scale converting, noise filtering for image enhancing is processed. After segmentation, calculate the eigenvectors of images and examined the associated word by using database. The various types of images are applied for training and testing. The top words are described for annotated image in result. Manual image annotation is time-consuming, laborious and expensive; so, there has been a large amount of research done on automatic image annotation and retrieval technologies are combined to improve the performance.