WARPIMAGEMULTITRANSFORM(1)

✨ NAME ✨

Okay bestie, meet WarpImageMultiTransform 🦋, a user command (manual section 1) that's part of the ANTS registration suite. It's giving image-warping royalty 👑.

---

✨ DESCRIPTION ✨

💅 Usage

Here's the basic vibe:

WarpImageMultiTransform ImageDimension moving_image output_image  -R
reference_image --use-NN   SeriesOfTransformations--(See Below)

🔗 SeriesOfTransformations

This is where the magic happens ✨. WarpImageMultiTransform can chain together an unlimited number of transformations on your data (concatenation, bestie!). So you could do an affine transform, then a warp, then another affine, then another warp. Iconic, honestly 🔥.

Inverse affine transforms get called with -i MyAffine.txt.

Inverse warps get called by passing the InverseWarp.nii.gz filename. Peep the important notes at the bottom about this 👀.

🧚 Example 1: Warped into the reference_image domain

Mapping a warped image into the reference_image domain by applying abcdWarp.nii.gz and then abcdAffine.txt:

WarpImageMultiTransform 3 moving_image output_image -R reference_image
abcdWarp.nii.gz abcdAffine.txt

🌈 Example 2: Back into the moving_image domain

To map the fixed/reference_image (warped) into the moving_image domain, you apply the inverse of abcdAffine.txt and then abcdInverseWarp.nii.gz:

WarpImageMultiTransform 3 reference_image output_image -R moving_image
-i  abcdAffine.txt abcdInverseWarp.nii.gz

Heads up! The inverse maps from Example 2 get passed in the reverse order of the forward maps from Example 1. Geometrically that totally makes sense 🤓. Check out ANTS.pdf for a visual of this syntax.

🫶 Compulsory arguments

• ImageDimension — 2 or 3, for 2D or 3D registration.
• moving_image — the image you want to apply the transformation to.
• output_image — the resulting image.

🌸 Optional arguments

• -R — the reference_image space you want to warp INTO.
• --tightest-bounding-box — figures out the tightest bounding box using all the affine transforms. If you pass -R reference_image, that overrides it.
• --reslice-by-header — same as -i -mh, or -fh -i -mh if you pair it with -R. It uses the orientation matrix and origin stored in the image file header. It can be used together with -R, but it's typically not used with any other transforms.
• --use-NN — nearest neighbor interpolation.
• --use-BSpline — 3rd order B-Spline interpolation.
• --use-ML sigma — anti-aliasing interpolation for multi-label images, with Gaussian smoothing at standard deviation sigma. Sigma can be in physical or voxel units, just like Convert3D. It can be a scalar or a vector.
  • Examples: --use-ML 0.4mm -use-ML 0.8x0.8x0.8vox
• -i — uses the inverse of the affine transform that follows it.

🎀 Other example usages

Reslice the image:

WarpImageMultiTransform 3 Imov.nii.gz Iout.nii.gz --tightest-bounding-box --reslice-by-header

Reslice the image to a reference image:

WarpImageMultiTransform 3 Imov.nii.gz Iout.nii.gz -R Iref.nii.gz --tightest-bounding-box --reslice-by-header

📌 Important notes

• A prefix name like abcd with no extension will use .nii.gz by default.
• The abcdWarp and abcdInverseWarp files don't actually exist. They're formed from abcd(Inverse)Warp.nii.gz when you call WarpImageMultiTransform, but you still have to use them as if they exist 💀.

---

That's a wrap, bestie! 🎀 Version 2.6.4... from December 2025, go slay those registrations 💖🫶