Real-time image registration method based on improved ORB algorithm for microscope

In recent years, there has been growing request in widening the field of vision of the microscope and stitching multiple images together into a panorama where the entire scene is visible in high-resolution. Considering the characteristics of microscope image, such as high complexity and variety of microstructure, previous researches always could not balance speed and accuracy of registration method. In this study, we creatively present a real-time microscope image registration algorithm by combining the phase correlation algorithm and the feature point matching algorithm. Firstly, the Gamma correction method was employed to preprocess the images to enhance their contrast. Then, the rough overlapping zones of the images were calculated by phase correlation. Thirdly, the improved ORB algorithm was used to extract the image features in the rough overlapping areas. Finally, the features were corresponded by matching algorithm and the distance between the images is determined, then the images were blended seamlessly. Experimental results show that the proposed algorithm has higher efficiency and precision compared with the traditional method.