Image Registration Algorithm Based On Hough Transformation

Image registration technology refers to the process to match two or more images on the same scene obtained at different time, different angle of view,or under different sensing methods, is a classical problem in image processing theory,has been widely used in computer vision, pattern recognition and medical image processing and other fields, Image registration methods are divided into two types:the registration method based on gray level and registration method based on the features, studies of this paper focus on the image registration method based on features.Image registration method based on features turns the registration problem of two global images into registration problem of feature points on two images to deal with, in order to reduce the time complexity of the algorithm, This idea decides that in this method, extracting and matching feature points on two images are particularly critical steps, because the mismatching of feature points will cause a big error on the result of registration transformation, although there are several kinds of matching algorithm for feature points have been verified with greater accuracy, but these algorithms are not applicable in image registration problem, only can be applied in the field of template matching and pattern recognition,cause they all take image spatial invariance as a premise.Therefore, on the basis of summarizing the predecessors’ research results, this paper puts forward a kind of image registration method based on Hough transformation, this method extends the traditional cross correlation matching algorithm to ensure the veracity of matching results of the feature points on images with arbitrary transformation relationship mostly, later eliminates the calculation errors caused by error matches with Hough transform when searching for the transformation parameters. The article first discuss the inner reasons for introducing the Hough transformation to improve the accuracy of registration, then demonstrates its strong robustness against the error matches, finally gives the detailed method to implement the algorithm, the experimental results, and all data that can illustrate the problems in the experiments, fully verified the effectiveness of the proposed algorithm.