ABSTRACT
The Content-based image re-ranking system retrieves the images by evaluating the visual features of the images. As there is a gap between visual concept and semantic meaning, the result does not show accuracy. To deal with this problem, the image re-ranking using semantic signature is used. The universal data set is reduced to semantic space, hence contains less images for computation. Also by calculating semantic meanings, it captures the user intension and gives accurate results. The duplicate images may occur in the re-ranked result. We have overcome the disadvantage of the image re-ranking using the semantic signature system by removing the duplicate images at the time of re-ranking. Also, we compact the semantic signature by using the hashing technique. It results in the improvement of matching efficiency. The re-ranking technique using hashed semantic signature shows better results than the re-ranking technique using semantic signature.
