Open nfaggian opened 13 years ago
Sounds good.
I like this idea because we can generate a mask based on the features. So cost function only measures cost around points of interest instead of the whole image.
Should work well with proposed SSIM metric as well, with SSIM being calculated for masked areas only.
We need to add masking to the registration process, minimizing:
e = | M(W(I;p)) - M(T) |
where M is a masking function defined in the template space...