users
users > Inverting, then exporting warp
Oct 25, 2014 01:10 AM | Torsten Rohlfing
Inverting, then exporting warp
Oh, and your other question - there is no tool to write the inverse
of a warp xform as an xform. That's because the specific class of
transformations, B-spline FFD, is not actually closed under
inversion. Which is to say, the inverse of an FFD is not in general
an FFD.
That said, you CAN use the "fit_spline_xform" tool to fit a B-spline warp to a series of arbitrary transformations. That includes the option of fitting to a single, numerically inverted warp. (Similarly, you can use fit_spline_dfield to fit a warp to a deformation field.)
But this fit is only an approximation, because as I said, there is in general no exact solution to the B-spline warp inversion problem, at least none that is itself a B-spline warp.
Also, the fitting may or may not work very well depending on what parameters you use for the fit (e.g., what control point spacing, and what, if any, multi-resolution fitting strategy you select.)
Again, hope this is helpful.
Torsten
That said, you CAN use the "fit_spline_xform" tool to fit a B-spline warp to a series of arbitrary transformations. That includes the option of fitting to a single, numerically inverted warp. (Similarly, you can use fit_spline_dfield to fit a warp to a deformation field.)
But this fit is only an approximation, because as I said, there is in general no exact solution to the B-spline warp inversion problem, at least none that is itself a B-spline warp.
Also, the fitting may or may not work very well depending on what parameters you use for the fit (e.g., what control point spacing, and what, if any, multi-resolution fitting strategy you select.)
Again, hope this is helpful.
Torsten
Threaded View
Title | Author | Date |
---|---|---|
Jaguar Li | Oct 24, 2014 | |
Torsten Rohlfing | Oct 25, 2014 | |
Jaguar Li | Oct 25, 2014 | |
Torsten Rohlfing | Oct 25, 2014 | |