Sub-pixel image registration method based on super-resolution reconstruction

Image registration is the process of geometric calibration of two or more images with overlapping areas of the same scene from different perspectives, different sensors or acquired at different times. This article focuses on the registration problem of sub-pixel displacement images in the process of super-resolution image reconstruction. Two sub-pixel image registration methods are introduced, including the spatial domain based feature matching SIFT registration method and the frequency domain based Vandewalle registration method. Four low resolution sequence images obtained through a micro scanning camera are used to simulate and analyze the registration accuracy of the two algorithms. The results show that both algorithms can achieve sub-pixel level registration of low resolution images.