Question

I transformed an image using meshgrid, but the new coordinates are partially outside the range of the original image, causing the transformation to fail.

First I used clip

...
tX = numpy.clip(tX.astype(numpy.int),0,image.w)
tY = numpy.clip(tY.astype(numpy.int),0,image.h)
result image([tX,tY])

which resulted in effect similar to the ‘nearest’ boundary condition.

I would like for all the outside pixels to be black. I thought I could achieve this by using a boolean array on the meshgrid, but I don’t know how to apply the boolean array to the meshgrid correctly.

tXbool = numpy.abs(tX) < image.w
tXbool.shape
(850, 1280)
tX[tXbool].shape
(193180,)

Answer 1

Okay, thats how I did it in the end:

tXbool = (0 < tX) & (tX  < image.w )
tybool = (0 < tY) & (tY  < image.h )
outliers = tXbool & tybool

mask = where(outliers,1,0)
mask4channels = dstack([mask,mask,mask])
image = image*mask4channels