Research on the registration of infrared and visible images based on phase consistency and edge extreme points

Abstract Due to the significant differences in the imaging principles of infrared images and visible images, the feature points and descriptors between the two cannot be effectively matched directly by traditional feature extraction methods such as SIFT. To solve this problem, this study proposes a registration algorithm for infrared and visible images based on phase information and edge information. The algorithm extracts the feature points of the infrared image and the visible image through the principle of phase agreement and the edge binary map of the image and then calculates the descriptors of the gradient images of the infrared image and the visible image, and the descriptor calculation draws on some SIFT principles. Finally, the cosine similarity was used to match the feature points, and the improved random sample consensus algorithm was used to screen out the correct registration points. Experiments show that this method can effectively register between infrared and visible images and is also suitable for the registration of infrared and visible images with different rotation angles and similar structures.