Efficient way of finding the minimum/maximum row index for each column in a numpy.array with nonzero value?

Question

I have a 2D numpy array iarr coming from a single color of a picture.

I want to find the minimum/maximum row index in each column with a nonzero value. If there are no nonzero values in a column this column doesn't need to be considered. I have a working solution but it is very slow. My current solution is this

img = Image.open('nameofimage.jpg')
iarr = numpy.array(img)[:,:,0]

nonz = numpy.nonzero(iarr)
colinds = numpy.unique(nonz[1])
minrowinds = numpy.array([numpy.min(nonz[0][nonz[1]==cind]) for cind in colinds])

Answer 1

Thanks to yatu's pointer, I can now answer this myself.

colinds = numpy.unique(nonz[1])

minrowinds = numpy.argmax((iarr>0),axis=0)[colinds]

For the maximum indices I had to flip the array first, as np.argmax always gives the first occurrence of the maximum value.

maxrowinds = numpy.argmax(numpy.flip((iarr>0),0),axis=0)[colinds]
maxrowinds = iarr.shape[0] - maxrowinds