A Survey Paper on To Retrieve the Relevant Images according to Re-Ranking of Images with Click based Similarity and Typicality

Searching performance can be improved with image re-ranking. Search engines provide the results according to top ranked set of images. The result set is processed and rearranged by using some specific features. Search engines like Google and Bing uses this concept of re-ranking. In these search engines initially textual query is processed and then pool of images are retrieved for the text query. A query image is selected from the pool and according to visual features of the query image, the pool of images are re-ranked. To get semantic signatures, the visual features are projected with semantic space specified by query. Proposed approach enhances matching efficiency of query specific semantic signatures. Similarities of visual features do not well correlate with image semantics. The visual features of images are projected into their related semantic spaces to get semantic signatures. At the online stage, images are re-ranked by comparing their semantic signatures obtained from the semantic space specified by the query. Thus proposed image re-ranking framework using query specific semantic signature improves both the accuracy and the efficiency of retrieval. A semantic signature is a list associated with an administered metadata object.