For those in the medical imaging community I've written some simple Matlab
software to do normalized cross correlation in three dimensions. For now, it's
a mere adaptation of Matlab's normxcorr2, and thus inherits many of its
shortcomings (namely, poor performance). My next step is to port it to C, and choose
suitable optimizations from the ones I list in my 2D NCC
tutorial. If you have input to give, I'd like
to hear it. For now, you can
download what I have:
January 7, 2006:
A much faster version that computes the numerator in the frequency domain.
normxcorr3.zip (zipped M-file)
December 12, 2005:
Horrifically slow software. Performed the NCC numerator computation in the spatial domain,
which is quite expensive except on uselessly small template/image sizes.
normxcorr3_spatial.zip (zipped M-file)
Back to my main page.
|