Re ranking system for query based facial-image retrieval

摘要

Query based facial-image retrieval is a common approach to retrieve the images subscribed by public users in social websites such as flickr. Anyhow Existing approaches for query based facial-image retrieval successively yield inappropriate, less diversity outcomes. In our work, we used Query based facial-image retrieval for re-ranking by examining the relevant images, Diversity and difficulty in computation, which aims to spontaneously ordering facial-images by using SIFT features, semantic features and views features. In the image search engine if the user gives an input query, images labeled with input query are retrieved as the primary outcomes. The primary outcomes contain facial-images subscribed by various users. First we sort the retrieved images by extracting semantic feature for all images. Image which has more semantic information will rank higher and it will selected as an query image. all relevant images will be retrieved by extracting SIFT feature. Then we implement relevance re-ordering on the ordered image collection, and also the more relevant images in the image dataset is elected. These elected images formulate the final obtained outcomes. We build an inverted index structure for fast retrieval and to improve the search operation.