Abstract: Copy-paste is a technique widespreadly used in image tampering,and it is one of the most covert tampering means.SIFT is a kind of common matching algorithm,it is also a more effective test method for the copy-paste tampering images.But it has the problems of poor matching accuracy,high time-complexity and so on.In order to overcome the problems,some improvements were made in this paper.The threshold value is determined by the method of fitting optimization to solve the problem of accuracy when the threshold is increased,and methods of extracting the feature points by SIFT algorithm are improved.The BBF search algorithm based on K-D tree is adopted to achieve the fast matching of the nearest neighbor query,and the feature matching is improved in SIFT algorithm,so the problem of high time complexity goes easy.The experimental results show that the proposed algorithms are effective.

Key words: Copy and paste forgery,Fitting method,SIFT algorithm,BBF search algorithm