Correspondence between intrinsic axes and world axes for imref objects

Suppose I have a 3D array of image samples I(i,j,k), i=1...M, j=1...N, k=1...P.

In world coordinates, pixel spacing in millimeters along the i,j, and k sampling directions are di, dj,and dk respectively. Now, I wish to create an imref3d object relating these samples to my world coordinates.

Question: According to the conventions of imref3d, are the world X, Y, and Z axes considered to be aligned respectively with increasing i,j, and k? Or is it j,i, and k like with meshgrid? Or something else entirely?

I am trying to determine whether, when using the construction syntax

R = imref3d(imageSize,pixelExtentInWorldX,pixelExtentInWorldY,pixelExtentInWorldZ)

the proper substitions of my parameters are as follows,

R = imref3d([M,N,P]  , di, dj, dk)

or something else? And what about imref2d? Any fundamental differences between the 3d and 2d conventions?

Answer 1

[Catalytic]

Or is it j,i, and k like with meshgrid?

It is this one, sad to say. You can check it like this

M=50; N=40; P=30;

ir = imref3d([M,N,P])

ir = 

  imref3d with properties:

           XWorldLimits: [0.5000 40.5000]
           YWorldLimits: [0.5000 50.5000]
           ZWorldLimits: [0.5000 30.5000]
              ImageSize: [50 40 30]
    PixelExtentInWorldX: 1
    PixelExtentInWorldY: 1
    PixelExtentInWorldZ: 1
    ImageExtentInWorldX: 40
    ImageExtentInWorldY: 50
    ImageExtentInWorldZ: 30
       XIntrinsicLimits: [0.5000 40.5000]
       YIntrinsicLimits: [0.5000 50.5000]
       ZIntrinsicLimits: [0.5000 30.5000]

As you see, the image extents in X,Y,Z are N,M, and P, so the usual meshgrid interchanging of X and Y is at work here as well :-(

[Matt J]

It is this one, sad to say

I tend to agree that it is sad, although I guess I also understand why some people like to have their X-axes run horizontally on the screen.