There's a problem with the border management in the morphological erosion. Currently we are doing:
Set all borders to 0, then work on all 3x3x3 windows. Thus avoiding to work with outside voxels.
This seems logical, but it always gives a black border, whatever the kernel used.
The consequence of our current method can be seen in the following image.
In red+pink we have ANTs ImageMath result.
in pink, we have ndimage result. We can easily see that we "ate" the border.
I readmore and I realized that it's not the right behavior: "Pixels beyond the image border are assigned the maximum value afforded by the data type." Thus all the padded voxels should be set to 1.
Our current method (working with windows to avoid padding) can't work. We can keep the window but we must pad with 1 if we want to avoid border effects.
There's a problem with the border management in the morphological erosion. Currently we are doing:
The consequence of our current method can be seen in the following image.
ImageMathresult.I read more and I realized that it's not the right behavior: "Pixels beyond the image border are assigned the maximum value afforded by the data type." Thus all the padded voxels should be set to 1.
Our current method (working with windows to avoid padding) can't work. We can keep the window but we must pad with 1 if we want to avoid border effects.