imrotate3 gives me just zeros with 'linear'

For imrotate3() the doc says:

"'linear' Trilinear interpolation. Trilinear interpolation is the default method for numeric and logical images."

Now if I have a 3D mask of zeros and ones... let's call it BW and it is of type logical, and I use

BW2 = imrotate3(BW,45,[0 0 1],'linear','crop');

Then BW2 is all zeros.

If I convert to, say, double or single, it works as expected.

Or if I use 'nearest' instead of 'linear', it works with logical.

Answer 1

[Jan]

I cannot confirm this, but it is even stranger:

BW  = rand(5, 5, 5) > 0.5;
BW2 = imrotate3(BW, 45, [0, 1, 0], 'linear', 'crop');

BW3 = 5×5×5 logical array
BW3(:,:,1) =

 204 160   0   0 158
  59 230   0 204 160
 127  63   0  59 230
 102   0   0 127  63
 158   0   0 102 105


BW3(:,:,2) =

 179 243   0   0   0
 125 191   0   0 240
  51   0   0   0  63
 219   0   0   0   0
 225   0   0   0   0


BW3(:,:,3) =

   0   0   0  59 230
   0   0   0 127 191
   0   0   0 102   0
   0   0   0 158   0
   0   0 203 160   0


BW3(:,:,4) =

   0 206 160 255  64
   0  59 230 255   0
   0 127  63 255   0
   0 102 254 255   0
 128 158 255   7   0


BW3(:,:,5) =

   0   0   0   0   0
   0   0   0   0   0
   0   0   0   0   0
   0   0   0   0   0
   0   0   0   0   0

ans = 
ans(:,:,1) =

     0     0     0     0     0
     0     0     0     0     0
     0     0     0     0     0
     0     0     0     0     0
     0     0     0     0     0


ans(:,:,2) =

     0     0     0     0     0
     0     0     0     0     0
     0     0     0     0     0
     0     0     0     0     0
     0     0     0     0     0


ans(:,:,3) =

     0     0     0     0     0
     0     0     0     0     0
     0     0     0     0     0
     0     0     0     0     0
     0     0     0     0     0


ans(:,:,4) =

     0     0     0     0     0
     0     0     0     0     0
     0     0     0     0     0
     0     0     0     0     0
     0     0     0     0     0


ans(:,:,5) =

     0     0     0     0     0
     0     0     0     0     0
     0     0     0     0     0
     0     0     0     0     0
     0     0     0     0     0

I;

BW2-BW3 assume this is worth to send a bug report to MathWorks. You've found the bug, so do you want to use the "Contact Us" link on the bottom of this page and the "Bug Report" on the following page?

[DGM]

One wouldn't guess by looking at it, but BW2 is 'logical' class. I would say that's quite the bug.

BW  = rand(5, 5, 5) > 0.5;
BW2 = imrotate3(BW, 45, [0, 1, 0], 'linear', 'crop');
class(BW2)

ans = 'logical'

It goes back to at least R2019b, fwiw.

[Mads]

I reported the bug.

I think it dates back even further.

[DGM]

If it was in R2019b, I'm going to guess it was present when imrotate3() was released -- R2017a. The IPT release notes don't mention any changes to it until R2020a, and there are no bug reports. Then again, the release notes aren't comprehensive and there are plenty of significant bugs that never get put in the publicly listed bug reports.