change camera parameters to field of view angles




stuff for muri demo:
	
	do probing with tiny skull and mikes pics
	reconcile sw and ogl rendering
	repair pose update normal calculation
	disable reg use of normal
	test reg
	enable reg use

	checkpoint code
	re-test video hardware usage



Clean up reg makefile





hack learning rate or optimization method

put in marks medical data io facility

debugging data is taking a lot of room.

---------------------------------------------------------------------------
STUFF TO FIX:					REMEDY


which random to use?

DESIGN


---------------------------------------------------------------------------


reg-2d-3d-i:
	hack for speed
	trim ron model
	tune params
		smoothness
		sigmas
		experiment:
			null pose stability
			easy delta pose, all parms
			capture testing

	better display?  edge overlay?

	prune back-facing normals
	different z learning rate

-------
central-slices.cc should fill in pixel sizes!


-------
general system stuff
	facilities to build
		landmark reg
		keyboard reg

	fortran numerical codes?
	look for string package
	sequence facility

general code
	makefiles
		the makefiles don't properly indicated the dependencies on
	general const correctness, and ref args when suitable
	flush form()
	flush <stream.h>
	INIT on strings sucks
	use drand instead of random()
	specialize instantiator names...
	be more private in class defs, use frendship more

etc code
	warnings about unused stuff in init files...

vec-mat code
	quat normalize becomes member

	printers
		print elt type in printer
		uniform field width
		better stream usage
		more error checking on stream printer and reader?
		colvec printing is BOGUS? just skip newlines if colvec?

	conditional range checking on access?
	do special transposes with inline
	consistent use of "return result"
	augment and reduce 
	for rmat: read functions...
	was lack of ref on m the problem in oveloading >> in mat?
	flush colveci , overload colvec?
	hack-pose:
		add initial pose capability
		add other order update for medical applications
		spit initial pose right away


image code
	gradients move to image and volumes	
	interpolators move to image and volumes
	images get pixel sizes
	more informative errors on type mismatch on reading volumes...
	volume readers don't fail when they should.
	uniformize window level setting syntax
	include ornt specs in volume files?
	swap hack
	the_volume.the_slices[i] is an ugly way to select slices.
	danger in copying images and volumes:
		leads to extra deletes !!! = BAD
		forbid it?  
		make it verbose?
		image "=" private?
	fill function


display code
	3 plane display
		geometric coords
		interactive plane selection
	gamma to work on one bit display
	floyd s. dithering
	get to work on 24 bit screen alphas...
	cruise-volume
		interactive slice motion
	cleaner colormap hacking...  less ad hoc, more tolerance for other ones



reg code
	1/t decay?
	reg progs to close displays
		print newlines after poses
		shut display and sync before writing poses

	estimate the entropy
	shift sucks











