Research On Image Registration Technology Based On Feature Points

With the rapid development of digital image processing technology,image registration,as a hot research direction in the field of image recognition,has a wide range of applications in the fields of medical image processing,remote sensing image processing,computer vision and face recognition.Image registration technology developed today,after several generations of scholars tireless research and exploration,in the field of image recognition has put forward a lot of excellent performance of registration algorithms.Among these registration algorithms,the scale invariant feature transform(SIFT)feature has good scale invariance and good tolerance to image rotation,scaling,illumination change and noise.Scale-invariant feature transform(SIFT)algorithm was proposed by Lower in 1999.After several years of improvement,the algorithm was finally summarized in 2004.However,with the rapid development of image technology,the resolution and capacity of the image that need to be processed are getting higher and higher,and the accuracy and real-time performance of the algorithm are also required.The limitations of the SIFT algorithm such as low efficiency have gradually emerged.Based on the basic principles of image registration,this paper firstly studies the mathematical theory basis of SIFT algorithm and the implementation process of the whole algorithm.In order to solve the problem of slow searching speed of feature points,a KD tree-based method of searching feature points was proposed,that is,the data set was divided layer by layer by KD tree structure,the index of feature points data set was obtained,and then the feature points needed for image registration were queried by the way of index,which improved the efficiency of data searching.On this basis,aiming at the problem of mismatching in the matching pairs after the initial registration,the improved random sampling consistency(RANSAC)algorithm is used to carry out the fine registration of the results of the initial registration,which effectively improves the registration accuracy.Finally,the improved SIFT algorithm is compared with the traditional algorithm,and the experimental results show that the improved algorithm is more effective.