search

pyimreg / python-register

Linear and non-linear image registration methods, using scipy and numpy.
Other
62 stars 35 forks source link

Implement CrossCorrelation metric #30

Open riaanvddool opened 13 years ago

riaanvddool commented 13 years ago

Based on this paper:

www.isprs.org/proceedings/XXXVII/congress/7_pdf/6_WG-VII-6/03.pdf

riaanvddool commented 13 years ago

Mm, not sure that the method described in the paper returns an error metric for all image pixels...

nfaggian commented 13 years ago

Window function on each pixel, often achieved by performing an fft.

stefanv commented 13 years ago

Just to make sure I understood the paper correctly: this is an iterative method?

They do not describe what happens for out-of-boundary pixels... could be quite important (it certainly is when minimising the sum of difference squared).

nfaggian commented 13 years ago

parameter search wrt to cc metric, not clear what to do with the correlation window on boundaries - I suspect we just mask out missing data or smooth out boundaries to fill in the gaps.