How to generate a real valued spectrum of a image?

I need to generate the real valued spectrum of an image. Therefore I mirroring the image (size MxN) at i=M and j=N to make it symmetric.

ImSize=10;
Image=peaks(ImSize);

Im_Vertical_Mirror= [Image; flip(Image(:,end-1),2)];
Im_Horiz_Mirror=[Im_Vertical_Mirror; flip(Im_Vertical_Mirror(1:end-1,:),1)];

fft_Im=fft2(Im_Horiz_Mirror);

This code does not achieve the desired solution. I think this is due to the index values which are not distributed symmetrically around zero. Therefore I tried:

fft_Im=fft2(circshift(Im_Horiz_Mirror,[ImSize-1, ImSize-1]));

Which does not achieve the desired solution as well. Imaginary part occure which are not caused by floating point errors, the imaginary parts are quite large.