Units of landmark coordinates

As it says in the title, I would like to know the “units” of the landmark points created by BigWarp. For example, if a point has coordinates (10,10,10), does that refer to 10 pixels in x, y, and z from (0,0,0), or would that be the physical units (in my case microns)? I ask because when plotting a 3d scatter plot of the landmarks, the moving and fixed landmark point clouds are drastically different sizes. This made me think that the units where in pixel space rather than physical (microns) space. However, when I multiply the coordinates by microns/pixel for x, y, and z, the point clouds are still clearly not in the same physical space. But, when I multiply only the moving space landmarks by microns per pixel and not the fixed space, the point clouds look about right.

Am I correct in assuming that the fixed landmarks are in physical units, while the moving landmarks are in pixel space? Or am I completely off base? Any help is much appreciated.

Hi @zmiller22,

All landmark points are in physical units defined by the “calibration” of your images.
Maybe check the pixel width/height/depth of your moving and fixed images?

Bigwarp assumes the units for moving and target are identical.
For example, behavior will be weird if the moving image has resolution of 1000nm and the fixed image has resolution of 1um.

I can try to give more concrete help if you show your point clouds and provide the calibration of your images.



Thanks for the quick reply. The images were taken using two different modalities with different resolutions. The moving volume (landmark points in red below) has voxel dimensions of 0.43x0.43x0.43 microns and is 1128x427x636 pixels. The fixed volume (landmark points in blue below) has voxel dimensions 0.6x0.6x0.3525 microns and is 700x700x287 pixels.

As you can see, the moving point cloud is much larger than the fixed point cloud when plotting just the raw landmark points. However, when plotted in terms of microns, the point clouds should be roughly the same size. If I could figure out what the “units” of each point cloud are, then I could multiply by the proper factor to convert to physical units. I need everything to be in terms of physical units because I ultimately want to apply the TPS transform generated from these landmarks to other point clouds where points are in physical units and compute various metrics (also in physical units). Thanks again for your help!

1 Like


Okay, my take away from the above is that the moving (red) points range from ~[0,1000] in x
(which looks like the pixel range), but the physical range should be: 1128 * 0.43 = 485.04 .

Is that the right interpretation?

I’m also asuming that the points your showing came out of the table exported from bigwarp.

If that’s the case, then bigwarp isn’t “seeing” your pixel resolution. If you’re running from Fiji using ImagePlus's - make sure the resolutions are set properly in Image> Properties. If you’re using xmls let me know, and I can describe what has to be in your xml to ensure the landmarks are in physical coordinates.


That is very helpful thank you. I did not create these landmarks, but looking at the image files it appears that the moving file does have the incorrect resolutions (set to 1x1x1). I am going to convert landmarks according to the proper dimensions and hopefully that does the trick. Cheers!

1 Like