Phase correlation based image alignment with subpixel accuracy

摘要: The phase correlation method is a well-known image alignment technique with broad applications in medical processing, stitching, and computer vision. This relies on estimating the maximum of phase-only (POC) function, which defined as inverse Fourier transform normalized cross-spectrum between two images. coordinates correspond to translation One main drawbacks this method, its basic form, that location can only be obtained integer accuracy. In paper, we propose new estimate subpixel accuracy, by minimizing magnitude gradient POC function around point near maximum. We also present some experimental results where proposed shows an increased accuracy at least one order respect base method. Finally, illustrate application algorithm rigid registration digital