Publication

Abstract : As the increase the number of images in the digital world makes the need for proper annotation for every image for proper identification. To facilitate the image identification major research has been carried out on automatic annotation to bring a common way of representation and unique identification in the huge image database. Because a single image may related to many label. It can be classified a search engine application also. A threshold multiple approaches can be applied to improve the searching process efficient and reduce the searching time. Annotation can be done with the help of semantic analysis comparison with text in the web page and probability string matching can be applied to search the images. Proposed approach involves the process of creating a graph with weights for each node and making a prediction for the images. Probabilistic approach reduces the number of miss hits by extended training.