Research On Image Retrieval Technology Based On The Scale Invariant Feature

With the rapid development of digital technology and Internet communications,digital images have penetrated into people's daily lives,such as traffic monitoring,health care and news media.And most of the digital images are in a state of disorder,the image of some important information can not be used,to find an efficient retrieval algorithm allows users to obtain the desired target image accurately and quickly,which has become an urgent problem.In view of the above requirements,this paper first elaborates the current retrieval algorithms,points out the shortcomings of the traditional retrieval methods,and emphatically studies the retrieval technology that Content Based Image Retrieval.It is shown that the local feature is more effective than the global feature to ensure the robustness of the image retrieval results.Therefore,this paper uses the classical Scale Invariant Algorithm as the image retrieval algorithm.However,this study found that the above algorithm there are more mismatched points,can not be practical applications and other deficiencies.So,this paper improves the algorithm,the first is to build the image feature descriptor,using the Best Bin First algorithm to improve the image feature point registration rate;the second is a combination of improved nearest neighbor matching algorithm and RANdom SAmple Consensus algorithm o further remove the wrong match point pairs.Finally,in a variety of experimental environments to experiment the improved algorithm.the results show that the improved algorithm can effectively reduce the error matching rate while preserving the advantages of the original algorithm,and can be used efficiently in the content-based retrieval system.This paper designs and implements a simple system based on the improved algorithm.The system is mainly composed of database establishment and database query.Through the comparison and analysis,it is shown that the improved algorithm can not only reduce the registration time in the actual retrieval system,but also improve the registration rate of the image.