How to use griddata from scipy.interpolate

Question

I have a three-column (x-pixel, y-pixel, z-value) data with one million lines. The data is from an image and there are duplicated z-values. Now I need to make a surface plot. This image is a perfect example. But now the output image is null. Could someone check the code please?

import numpy as np
from enthought.mayavi import mlab
from scipy.interpolate import griddata
x,y,z = np.loadtxt('test.csv',delimiter=',',usecols=(0,1,2),unpack=True) 
xi,yi = np.mgrid[0:3000:3000j, 0:3000:3000j]
zi = griddata((x, y), z, (xi, yi),method='linear')
mlab.surf(xi,yi,zi)
mlab.show()

Answer 1

I can't check the code without having the data, but I suspect that the problem is that you are using the default fill_value=nan as a griddata argument, so if you have gridded points that extend beyond the space of the (x,y) points, there are NaNs in the grid, which mlab may not be able to handle (matplotlib doesn't easily). Try setting fill_value=0 or another suitable real number.