Using a user defined fucntion with blockproc

Is there a mistake in my syntax using blocproc:

   Binary1=blockproc(OrigImage,[tileRowSize, tileColSize],@(x) myOperation(hObject, eventdata, handles, x.data));

the function I use "myOperation" is as follows:

   function myOperation(hObject, eventdata, handles,tile)
      [thresh1, thresh2] = Threshold(hObject, eventdata, handles,tile);  %my personal thresholding function

      %Create Binary Image
      size(tile)
      ix=tile;
      level=thresh2
      ix(ix<level)=0;
      ix(ix>level+1)=1;

Im getting

Binary1 =

       []

Answer 1

[Image Analyst]

I don't have time to look at it now, but in the meantime, see my blockproc demos, attached.

Answer 2

[Iain]

What does your "myfunction" function output? It looks like it's not producing any output.

[Jason]

Nothing, should it be?

Im guessing of course it should, I have now changed :

function [BI] = myOperation(hObject, eventdata, handles,tile)

and it works. Thankyou

[Jason]

Now I have the blockproc approach thanks to everyones kind help, I can see by comparison to the subdivision using looping (as initially suggested by IA) is almost the same , but not quite!

Comparing the two binary images obtaine dby these two approachs with the raw image in the highlighted boxes I can see that consideration of the additional row/column via blocproc seems to be falsely picking out objects.

I also notice using the blockproc approach (my raw image is 101x101, and Im using 25x25 in blockproc), that the very last pixel is the actual grayscale and not binarised?

      0      0      0      0      0
      0      0      0      0      0
      0      0      0      0      0
      0      0      1      1   1914

[Image Analyst]

If you have blocks of 25 pixels by 25 pixels, and you move in "jumps" of 25 pixels, then I believe your last block will have only 1 row or column. Thus it may not have a value like the others that had a full block in them.

[Jason]

Yes that makes sense.

Is it possible to return also the mean intensity of each sub image using blockproc, I thought I could just add an extra output variable from my function such as:

function [BI, mn] = myOperation(hObject, eventdata, handles,tile)

   mn=mean(tile)
  %then the rest of the code.

and then obtain a matrix of mean values via:

[Binary1, mn]=blockproc(OrigImage,[tileRowSize, tileColSize],@(x) myOperation(hObject, eventdata, handles, x.data));

But it complains about too many output arguments for Blockproc.