====== Resampling vs Registration ======

Both resampling and registration (also spatial normalization/warping) modify the underlying matrices of one image to match that of another. But unlike resampling, only registration and warping inspect and modify based on the data within the images.

<WRAP=tip>Resampling is concerned with the matrix shape, registration with the matrix content.</WRAP>

From [[https://andysbrainbook.readthedocs.io/en/latest/FrequentlyAskedQuestions/FrequentlyAskedQuestions.html#resampling|Andy's Brain Book]]

> Resampling means changing either the resolution, dimensions, or both the resolution and dimensions of an image.

Compared that to [[https://andysbrainbook.readthedocs.io/en/latest/fMRI_Short_Course/Preprocessing/Registration_Normalization.html#registration-and-normalization|warping]] where
> each brain [within an image] needs to be transformed

===== 3dresample vs flirt =====

{{:mr:pasted:20241124-173608.png|resample vs affine transform}}

''%%3dresample%%'' is an AFNI tool to resample an inset to the "master". We can compare it to FSL's ''%%flirt%%'' which reorients a brain to a reference (''%%-ref%%''). Both commands produce a new image with the same grid. However, the output of the ''%%flirt%%'' command will have a brain that's been moved. The value of the voxel at ''%%(i,j)%%'' ''%%(1,1)%%'' will be different in the input, resampled (''%%3dresample%%''), and registered (''%%flirt%%'') images.