PROJECT TITLE:

EMR A Scalable Graph-based Ranking Model for Content-based Image Retrieval - 2015

ABSTRACT:

Graph-based ranking models are widely applied in data retrieval space. In this paper, we tend to specialise in a well known graph-primarily based model - the Ranking on Knowledge Manifold model, or Manifold Ranking (MR). Particularly, it has been successfully applied to content-based image retrieval, as a result of of its outstanding ability to find underlying geometrical structure of the given image database. However, manifold ranking is computationally terribly expensive, which significantly limits its applicability to giant databases particularly for the cases that the queries are out of the database (new samples). We tend to propose a novel scalable graph-based mostly ranking model known as Efficient Manifold Ranking (EMR), attempting to deal with the shortcomings of MR from two main perspectives: scalable graph construction and economical ranking computation. Specifically, we tend to build an anchor graph on the database instead of a traditional k-nearest neighbor graph, and style a new kind of adjacency matrix utilised to hurry up the ranking. An approximate technique is adopted for efficient out-of-sample retrieval. Experimental results on some giant scale image databases demonstrate that EMR is a promising technique for universe retrieval applications.