

Notes

neurospin/registration

registration/

 __init__.py
	registration.py
	iconic_registration (intensity based, joint histogram)
	 renamed joint registration
	 takes from and two images  and compute joint histogram
	groupwise_registration.py (motion correction in fmri)
		register a set of images
		sum of square differences
		not using joint histogram
	affine.py (discribes a general 3d affine transformation and its parametrization)
	 class affine
		  params=s(-1)xv12 s: pre_cond
		  radius for the preconditioner is in translation coordinates
		  check for rigidity
	 class 
		  
	grid_transform.py (discrete displacements of the from grid)
	cubic_spline.c (same results as ndimage)
	wichmann_prng.c (only for the random interplation)
	iconic.c to be renamed to histogram.c
	
				  
	    

interpolating the histogram 
avoids the problem of casting the intensity

in C assumes the joint histogram is a signed short array (16bit)

clamp

Make independent tests with checks starting from different registrations.
Sensible default for the focus function
What should we do when outside the fov?




