Image Registration by Curvature Shape Representation and Genetic Algorithm

A new feature point extraction method for the image feature point matching is proposed. The proposed method is based on the corner detection method with curvature scale space (CSS). This method can accurately extract the image corner points in different positions and directions. In order to accurately match the corner points of two images, an overall restricted condition, which combines angle difference, gray level difference, relative distance and normalized correlation coefficient of the two matched corner points, is used to improve the matching accuracy. Finally, genetic algorithm is used to obtain the optimal registration parameters. The optimal registration parameters are used to accurately match the two images. The experimental results show that the proposed method can accurately match the images and better than traditional image registration method.