Convolution to model astigmatic aberration using zernike polynomials

Currently I am trying to use a PSF generated from the zernike polynomial for astigmatic abberation to model what an image I have would look like with astigmatic abberation.

Here is my code currently:

clear;

gcp;

name = 'AVG_dendrites.tif';

dendrites = imread(name);

imshow(dendrites);

N = 2;

M =2;

L = size(dendrites,1);

X = -1:2/(L-1):1;

[x,y] = meshgrid(X);

x = x(:);

y = y(:);

[theta,r] = cart2pol(x,y);

is_in_circle = ( r <= 1 );

E = zernfun(N,M,r(is_in_circle),theta(is_in_circle));

psf = abs(fftshift(fft2(ifftshift(E)))).^2;

psf = psf/sum(psf(:));

How do I create a kernel from this PSF so that I can convolve with the AVG_dendrites image to create the new astigmatic image?

[Bjorn Gustavsson]

This sounds like a dodgy idea. What type of astigmatic aberration is shift-invariant over the focal plane? We can only model shift-invariant point-spread-functions with convolution, and all aberrations I've heard about varies with distance from the centre of the image.