Research On Image Matching Based On SIFT Algorithm

Image matching is an important research topic in the fields of computer visionand image processing．The purpose of Image matching is to find similar couplesbetween two or more images by using a certain matching algorithm，the matchingimages are from the same object，often based on different time，different perspectivesor different viewpoints．Image matching methods cover two categories：one is based ongray，the other is feature．The former is based on gray value of matching images，butthe calculation is very large；The latter is not sensitive to noise，and can be used in thereal-time match．The key problems focus on how to increase the algorithm speed，to increase theregistration precision and to enhance the robustness of these registration methods．So，It is a great challenge to investigate a robust，precise，stable，and applicable stereomatching algorithm，which is both theoretically and practically valuable．A great deal of work is done to study image feature extraction and matchingalgorithms in this paper．Firstly，two typical image matching algorithms are described，Including：Harris Corners Detection and SIFT feature matching，introduced the basicideas of these two algorithms，their own implementation and process．Then newimproved algorithms are given based on the proposed algorithms：1.With the Harrisoperator to filter the Scale-space extreme points，access to high-contrast features；2.According to the circle rotation invariance，using a circular neighborhood of featurepoints to describe the point；3.Reduction the dimensionality of keypoint descriptorbased on the combined gradient；4.Taking two-way matching strategy to improve theaccuracy of matches；Finally，I described the basic ideas of improved algorithms，theprocess of its implementation，and use experiment to analyze the matching results todiscuss these algorithms．...