With thecontinuous development of the Internet technology and the popularizationof digital products,various images and videos are uploaded to the Internet,which still keep growing at an astonishing speed.It has been a key problem that how to quickly search the relevant information among these huge datasets.Our work is focused on the content based large scale fast image retrieval.Many classical content based image retrieval methods are focused on the extraction of visual information of image,such as color,texture and local key point features.However,these classical methods usually cannot achieve good performance for the increasing large image datasets,which is because of the restrictions of one single feature.According to the above situation,we proposed a hashing based large scale image retrieval(HLSIR)algorithm for achieving real-time retrieval.The proposed algorithm can simultaneously learn the compact hash codes and hash functions by combining global and local features.The efficient Hamming distance and bit count operations can be used to compute the similarity of two images,based on which the relevant images are retrieved.The efficiency and effectiveness of the proposed HLSIR method are examined on two public image collections.The results of extensive experiments show the superior performance of the proposedmethod over various classical and state-of-the-art algorithms for image retrieval. |