Semantic-Based Image Retrieval by Visual Analytics Using Ranking Algorithm

Abstract

Now-a-days owing to syntax attributes and ability of compression of vast data, CBIR techniques are proving for useful. Semantic-based image retrieval (SBIR) is a retrieval technique based on the intuition of user most SBIR algorithms is explain and retrieved images are considered relevant to the user query, there is a chance that the user may miss out relevant images which may be outliers. Semantic technologies like ontology offers auspicious approach to image retrieval as it tries to map the low level image features to high level ontology concepts. So a tool called Visual Analytics Tool for Semantic Retrieval (VATSR) is proposed which allows user to visually interact and refine the query and/or search results. Our approach makes full use of the explanation in query sketches and the top ranked images of the initial results. Relevance feedback is applied to find more relevant images for the input query sketch. Color and texture features are normalized. Finally Ranking algorithm is used to rank the images for the order of retrieval. Benefits of our proposed system applied in flickr are experimentally shown in terms of both relevance and speed.