v3regrb − Three-dimensional Rigid-Body Registration

v3regrb [-H] [if=<inputfile>] [ig=<inputfile2>] [of=<outputfile>] [df=<value>] [m=<value>] [t=<value>] [r=<value>] [dt=<value>] [dr=<value>] [-ex] [-h] [ha=<value>] [-p] [pt=<value>] [-v] [-nn] [-t] [ic=<value>] [-2d] [-com] [mf=<value>] [mg=<value>] [-jh] [wg=<value>]

v3regrb registers two 3-d images using a variety of metric functions and search algorithms. The output is the first image registered to the second image. The available metric functions are Mean-Squared-Error, Correlation, Mutual Information, and Normalized Mutual Information. The available search algorithms are Exhaustive, Hill-Climbing (Greedy), and Powell’s method. Registration information is dumped into the output file history.

if=

input file: image 1

ig=

input file: image 2

of=

output file: image 1 registered to image 2. The output image will have the same bounding box as image 2.

df=

the difference image of the registered image and image 2

m=

the metric to use: 0=Mean-Squared Error, 1=Correlation, 2=Mutual Info, 3=Normalized Mutual Information. The default metric is Mean-Squared-Error.

t=

tx,ty,tz : the range of translation for hill-climbing and exhaustive search. The default is half the bounding box.

r=

rx,ry,rz : the range of angles (degrees) for hill-climbing and exhaustive search. The default is 10 degrees.

dt=

the translation stepsize for hill-climbing and exhaustive searches, or the tolerance (minimum change in translation) for Powell’s method. (default=1)

dr=

the rotation stepsize for hill-climbing and exhaustive searches, or the tolerance (minimum change in rotation) for Powell’s method. (default=1)

-ex

Register using an exhaustive search (might take a while)

-h

Register using a hill climb search. This is a greedy search that steps in the best direction.

ha=

Hill Climbing parameter: Number of scaled passes. This will perform successive hill-climbing searches with a decreasing stepsize. (default=1)

-p

Register using Powell’s method (see numerical recipes)

pt=

Powell’s method parameter: metric tolerance, the minimum change in the metric function for the algorithm to exit. (default=0.1)

-v

Verbose mode.

-nn

Use nearest neighbor interpolation instead of tri-linear interpolation

-t

Register using translation only.

ic=

tx,ty,tz[,rx,ry,rz[,px,py,pz]]: the initial conditions -- translation, rotation, and rotation point

-2d

Perform 2d registration.

-com

Use center of mass to find the initial conditions.

mf=

image 1 mask - the metric will only be calculated for pixels in this mask

mg=

image 2 mask - the metric will only be calculated for pixels in this mask

-jh

Calculate the joint histogram between the two images and exit.

wg=

x,y,z,r: Apply a weighted gaussian to the metric function. The gaussian is centered at (x,y,z) and has a standard deviation of r. This is useful for registering juxtapleural and pleural-tail nodules.

Can only handle byte, float, or short images.

--

v3reg(1), v3regrb(1)

A. B. Chan, Summer 2001