Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > sci.image.processing > #4061

comparisons between phase correlation vs normalized cross correlation for image registration

Show all headers | View raw

Hi all,
I start learning about image registration and try to think about the limits/advantages of phase correlation compared with normalized cross correlation (NCC) for image registration.

One of the disadvantages of phase correlation I think is that phase correlation may perform worse than NCC in case we try to align a small template (moving image) with a big reference image (fixed image). The reason is the method calculates Fourier transforms of the whole reference image. The frequencies belong to the non-overlapping area (between the template and the reference image) will impact the phase correlation map. We may not get a sharp peak for this case.

NCC in the other hand is calculated and normalized in the overlapping area only and in this case the features that belong to the reference image and are not in the overlapping area (between the template and the reference image) won't affect the NCC score.

What do you think about my thought?

Also do you have any other ideas to share?

Thank you

Back to sci.image.processing | Previous | Next | Find similar

Thread

comparisons between phase correlation vs normalized cross correlation for image registration MBALOVER <mbalover9@gmail.com> - 2015-11-17 16:34 -0800

csiph-web