Specific query semantic signatures in web page re-ranked image retrieval

Index-based image retrieval improves results of a web image search in the customized real-time image search engines. Index-based Image ranking is an efficient method to improve image search results wherein, a user presents a query in customized search engine to retrieve a large number of images from the image pool on the basis of textual information. Traditionally, Index-based hash code image retrieval technique have been used to retrieve images with indexed hash codes that were arranged based on hash descriptors. In view of the presence of hash codes, re-rank-based image retrieval is not possible in index-based hash code image retrieval. Due to different visual features of images, re-ranking is not possible in index-based hash code image retrieval. An efficient and effective approach is needed to process re-rank with index-based image retrieval that is based on different visual features. Hence, in this paper, propose to implement and develop a Novel Image Re-Positioning System (NIRS) that automatically learns offline the various visual semantic features for different queries through keyword expansions. This novel approach significantly improves both accuracy as well as efficiency in image re-ranking for efficient image retrieval. Our experimental results show that relative improvement has been achieved on re-ranking with state-of-the-art methods.