Iterative Phase Correlation Algorithm for High-precision Subpixel Image Registration

článek v časopise ve Web of Science, Jimp

en

Many astrophysical observations and measurement techniques that rely on data from images include an image registration step. The results of such techniques thus heavily rely on the precision of the registration. We present an Iterative Phase Correlation (IPC) algorithm, which is an extension of the well-known phase correlation method of image registration and is ideally suited for problems, where the subpixel registration accuracy plays a crucial role. Furthermore, a sophisticated and reliable method of optimal IPC parameter estimation is described. The paper includes examples of such optimized parameters for Solar Dynamics Observatory (SDO)/Helioseismic and Magnetic Imager, SDO/Atmospheric Imaging Assembly, and Solar Terrestrial Relations Observatory A/B Sun Earth Connection Coronal and Heliospheric Investigation images. The new method (both with or without the parameter optimization step) significantly outperforms standard image registration methods, such as (non-iterative) phase correlation or (normalized) cross correlation in the sense of subpixel accuracy. A step-by-step pseudocode implementation is also included.

Many astrophysical observations and measurement techniques that rely on data from images include an image registration step. The results of such techniques thus heavily rely on the precision of the registration. We present an Iterative Phase Correlation (IPC) algorithm, which is an extension of the well-known phase correlation method of image registration and is ideally suited for problems, where the subpixel registration accuracy plays a crucial role. Furthermore, a sophisticated and reliable method of optimal IPC parameter estimation is described. The paper includes examples of such optimized parameters for Solar Dynamics Observatory (SDO)/Helioseismic and Magnetic Imager, SDO/Atmospheric Imaging Assembly, and Solar Terrestrial Relations Observatory A/B Sun Earth Connection Coronal and Heliospheric Investigation images. The new method (both with or without the parameter optimization step) significantly outperforms standard image registration methods, such as (non-iterative) phase correlation or (normalized) cross correlation in the sense of subpixel accuracy. A step-by-step pseudocode implementation is also included.

phase correlation, optimization, sub-pixel registration

20.02.2020

IOP PUBLISHING LTD

BRISTOL

0067-0049

247

1

247–255

8