Comparative Study on Automatic Image Annotation

With the detonative development of internet technologies in the web huge amount of images are available on the web. Large amount of research has been carried out on image retrieval since last few years. There is need for efficient and viable procedure to find visual information on demand. Recent research shows that there is semantic gap between low level features of image and semantic concepts understood by the humans. Handling large volumes of digital information become vital as online resources and their usage continuously grows at high speed online image sharing applications are getting extremely popular. This shows the potential of these online image collections and the need to search them on the basis of words. So to enhance the image retrieval results annotation is required as label associated with image represents the semantic information associated with it. The traditional approach for assigning relevant keywords to image is manual annotation in which images are annotated manually by humans but the manual annotation is labor intensive and time consuming. To overcome this problem and to bridge the semantic gap between low level concepts of image and semantic concepts understandable by humans approach is presented that automatically annotated the image which is called automatic image annotation. In this paper we have focus on various approaches for automatic image annotation. The analysis and key aspects of automatic image annotation is also presented. This paper also aims to cover automatic tagging challenges and future directions related to it. Keywords— Automatic Image Annotation (AIA), Content Based Image Retrieval (CBIR), Probabilistic Model, Support Vector Machine, Artificial Neural Network, Gaussian Mixture Model