An efficient subpixel image registration based on the phase-only correlations of image projections

This paper presents an efficient method of subpixel image registration based on the phase-only correlations (POCs) of image projections. Conventional POC-based methods need the two-dimensional discrete Fourier transforms three-times for calculating the 2-D POC of images. In order to reduce the computational cost caused by the two-dimensional discrete Fourier transforms, we decompose the two-dimensional POC of images into two one-dimensional POCs of the image projections. The vertical displacement and horizontal displacement are estimated with subpixel accuracy using the POC of the horizontal image projections and the POC of the vertical image projections, respectively. Experimental results show that the proposed method can estimate the subpixel displacement between two given images with high accuracy at a low computational cost.