Research Of Image Registration Based On The Feature Points

Image registration is a key technology in the field of image processing and pattern recognition. This technology is widely used in the field of computer vision, remote sensing image processing, medical image analysis, etc. Image registration based on feature points is the most widely used type of algorithm. But with the requirement of engineering application is increasing, the existing real time and accuracy of image registration algorithm are facing challenges. This paper first discusses the basic theory and principle of image registration, elaborates detection operator of common feature point and make a comparative analysis of its performance.This article focuses on the SURF operator studied. Its correct matching rate is low because it only takes into account the information of point descriptor, while ignoring the relationship between the feature points. Combined with gray difference histogram and relationship between the feature points, the SURF algorithm has been improved. The results of this paper show that correctly matched rate of feature points of the improved algorithm compared with existing algorithms improved significantly, and maintaining a high time efficiency of SURF algorithm.