Designing Image Search Engine Based On Attribute Assisted Re-Ranking Model

Information Technology

The huge growth of digital images over web, required the best image retrieval methods that can improve the retrieval accuracy of the images. Therefore research focus has been shifted from designing of refined algorithms that can reduce the semantic gap between visual features and richness of human semantics. Therefore many image re-ranking techniques has been developed to enhance the text based image results by taking the advantage of visual information contained in the images. But the previous techniques are based on low level visual features. Hence in this paper the semantic attributes and low level features are exploited simultaneously by using hypergraph re-ranking method. Based on classifiers for all the predefined attributes, each image is represented by an attribute feature containing of the answers from these classifiers. A hypergraph model is the association between the images and its relevance score to order the images. It’s simple based on that visually similar images should have related ranking scores. This modelling link among more close samples & will be able to domain the robust semantic similarity, hence expedite the great ranking performance. We conduct trials on more than 1000 queries in MSRA-MMV2.0 data set. The experimental results show the success of these methods.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.