Survey on Web Image Re-Ranking Using Query-Specific Semantic Signatures

In todays world accessing of an image using search engine is a important task, but getting a relevant image is a complex task. To improve the result of web-based image search image re-ranking is effective way. In web based image search like google, bing use text query or query to retrieve the images. From this pool of images user select one image as query image and remaining images re-ranked based on their visual similarities. A main contend is, the similarities of visual features do not well correlate with semantic meanings of images which understand users search purpose. Commonly used social multimedia websites like Photo bucket, flicker, Amazon is used to retrieve images in such a large network is very useful but also very monotony or challenging process because there exists lots of information such as images, text and network structure. It also takes more time and the retrieved contents are not exactly always same. novel internet image search require user only click on one image as query and images from the remaining images retrieved by text based search are re-ranked based on visual and texture feature. In Novel image re-ranking framework, offline learns different visual semantic spaces for different query s through expansions. At the online stage, images are re-ranked by comparing their semantic signatures obtained from the visual semantic space specified by the query.