Indian Journal of Science and Technology

Abstract

Objective: The main objective of this research is to reduce the semantic gap between human-understandable high-level semantics and machine generated low-level features for Automatic image annotation in Online Image Retrieval system. The semantic gap reduction is done by integrating the ontological concept with the Semantic annotated Markovian Semantic Indexing. Methods: The image annotation is the plays major task in the online image retrieval system by retrieving the user required images. In the existing system Latent Semantic Indexing (LSI) and Markovian Semantic Indexing (MSI) methodologies are used for the online image retrieval. However this work cannot retrieve the images in the accurately due to lack of semantic knowledge about the user submitted high level key words. This issue is resolved in the proposed research methodology by introducing the technique called Semantic annotated Markovian Semantic Indexing (SMSI) which is used for retrieving the images and automatically annotates the non-annotated images in the database using hidden Markov model. In contrast to traditional annotation based image retrieval system which retrieves images based on low-level features, the proposed SMSI semantically retrieves the images by searching semantically annotated images in a database for a user query. Each non-annotated image in a large collection of training samples would be annotated automatically with a posteriori probability of concepts of annotated images present in it. At last semantic retrieval of images can be done by measuring semantic similarity of annotated images in the large database by using Natural Language processing tool namely WordNet. In addition to that entity based ontology representation is introduced which tend to reduce the semantic gap between the human defined higher level keywords and machine specific lower lever features. Results: The presented SMSI method possess definite theoretical advantages and also to achieve better Precision versus Recall results when compared to Latent Semantic Indexing (LSI) and Markovian Semantic Indexing (MSI), methods in Annotation-Based online Image Retrieval system. The better accuracy is achieved while retrieving the contents based image annotation where the semantic gap is reduced considerably. Conclusion: Thus the analysis of presented work is demonstrates semantically related features of images and achieves improved retrieval result when compare with the other state-of-art techniques.
Keywords: Automatic Image Annotation, Latent Semantic Indexing, Markovian Semantic Indexing, Semantic Annotated Markovian Semantic Indexing