Fast Image Retrieval Based on Multi-feature LSH Indexing
Jun Feng · Journal of Shanxi University · 2013
With the development of satellite remote sensing technology,the earth observation data increases sharply.How to query and retrieve the remote sensing image automatically and effectively has become an urgent issue.LSH is now one of the best multidimensional database indexing techniques.It can effectively solve the curse of dimensionality and retrieval time consuming problems.To satisfy the practical requests of a content-based remote sensing image retrieval system and improve its retrieval efficiency,this paper proposes a new method that combines feature selection and LSH indexing validation.Borrowing ideas from feature selection,we introduce a new index to evaluate LSH indexing in different feature spaces and thus select the optimal features which can describe the retrieval objectives best.The experimental results on various land cover retrieval in several remote sensing photographs prove the efficiency and effectiveness of the proposed method.