Import of 3D affine transformation build in Java to R

Question

I have a question related to affine transformations (between the <affine> tags). I am extracting the following two affine transformations from an xml file (full xml file here) created with the BigDataViewer Fiji plugin created with Java, using the AffineTransform3D function:

<ViewRegistrations>
<ViewRegistration timepoint="0" setup="0">
  <ViewTransform type="affine">
    <Name>Fast 3d geometric hashing (rotation invariant), AffineModel3D on [beads (c=0)]</Name>
    <affine>1.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 1.0 0.0</affine>
  </ViewTransform>
  <ViewTransform type="affine">
    <Name>calibration</Name>
    <affine>1.9998662296836334 0.0 0.0 0.0 0.0 1.9998662296836334 0.0 0.0 0.0 0.0 1.9998662296836334 0.0</affine>
  </ViewTransform>
</ViewRegistration>

I would like to import the two affine transformation in R, using the buildAffine() function from the R package {RNiftyReg}, then compute their composition using composeTransforms() from {RNiftyReg}.

buildAffine(translation = c(0, 0, 0), scales = c(1, 1, 1), skews = c(0, 0,0),
  angles = c(0, 0, 0), source = NULL, target = NULL,
  anchor = c("none", "origin", "centre", "center"))

My question:
The affine transformations above are stored in a vector of 12 indices. buildAffine() requires as input parameters the values for the translations, scales, skews and angles.
I would like to know which value correspond to what.

The doc for the R package is indeed self-documented, but not the help for the java function which only says: `AffineTransform3D(double xx, double yx, double zx, double tx, double xy, double yy, double zy, double ty, double xz, double yz, double zz, double tz)`. So I am looking if anyone deals with java to give a real name to these parameters. — Marion, Oct 14 '18 at 20:06

score 0 · Accepted Answer · answered Oct 14 '18 at 23:11

I'm mostly an R user but here goes: The names of the varaibles in hte Java call are:

 dat <- scan (text="double xx, double yx, double zx, double tx, double xy, double yy, double zy, double ty, double xz, double yz, double zz, double tz", what="")
dat <- dat[dat != "double"]
matrix(dat,4)
      [,1]  [,2]  [,3] 
[1,] "xx," "xy," "xz,"
[2,] "yx," "yy," "yz,"
[3,] "zx," "zy," "zz,"
[4,] "tx," "ty," "tz"

This page documents how an affine transformation could be encoded with the only difference being that the 4x3 matrix is transposed: https://o7planning.org/en/11157/javafx-transformations-tutorial

 t( matrix(dat,4) )
     [,1]  [,2]  [,3]  [,4] 
[1,] "xx," "yx," "zx," "tx,"
[2,] "xy," "yy," "zy," "ty,"
[3,] "xz," "yz," "zz," "tz"

So the matrix would be applied to a vector of c(x,y,z,1) to get an output of:

Import of 3D affine transformation build in Java to R

1 Answers1