Research On Image Geometric Registration Algorithm

Image geometric registration algorithm is mainly studied in this paper.In this paper, firstly,the concept of image geometric registration and the research background of image geometric registration and the related technology are introduced, then the image geometric registration algorithm is introduced. The image geometric registration algorithm can be divided into gray registration algorithm,feature registration algorithm and transform domain registration algorithm. The main steps of the three kinds of registration algorithm are introduced in this paper, and the advantages and disadvantages of each main link are pointed out. After the main links of each class of algorithms are introduced, the common registration algorithm of each method are introduced, and the performance of the algorithm are verified by experiments.When the SIFT image registration algorithm and SURF image registration algorithm are used to deal with the problem of large rotation between floated image and reference image under the condition that no error registration algorithm is used,the registration accuracy is not high. To solve this problem,the image feature registration algorithm based on Fourier Merlin transform is proposed in this paper. Fourier Merlin transform is used to obtain the rotation angle difference between the floated image and reference image in this paper,then the floated image can be modified by the obtained angle,and the preliminary registrate image is obtained, the preliminary image are registrated by the feature registration algorithm.Experiments show that,the registration which there is large rotated between floated image and reference image can be effectively processed by the image feature registration algorithm based on the Fourier Merlin transform.Aiming at the problem that one-to-many feature association occured in SURF algorithm, resulting in a decrease in registration accuracy, Super-SURF image geometric registration algorithm is proposed in this paper.In this paper, the information richness region is extracted, and be used to detect feature points, then the better feature point pairs are selected by the method of selecting associate point pairs by sorting which is proposed in this paper,then the selected point pairs are used to calculate the transformation matrix,finally,the registration image is obtained.Experiments show that Super-SURF image geometric registration algorithm has a higher registration precise than that of SIFT algorithm and SURF algorithm, and its running time is less than SIFT algorithm and SURF algorithm.