Study Of The Algorithms For Image Matching

Image matching is an important research topic in the fields of computer vision and image processing. A great deal of work is done to study image matching algorithms in this paper.The two methods of image matching which are the method based on the lightness information directly and the method based on the feature of the images are researched mainly in this thesis. After the classic algorithms is analyzed and evaluated, the improved method of image matching is summed up.On the base, the new composite matching algorithm based on image geometric feature and transform domain correlation is proposed in this thesis by combining the algorithm of matching in frequency domain with image feature extraction effectively. The experiment result shows that this method can deal with the images existing shift and rotation well.For the requirement of making use of image registration in the system of danger detection, the matching algorithms between visible-light image and millimeter-wave image is put forward, which is proved effective between certain limits. Accordingly, This method can offer the guarantee for discovering dangerous object rapidly, also can enlarge the application field of the image matching technology.