The following code block shows how the SciPy signal module's fftconvolve() function can be used to run the convolution in the frequency domain (internally by just a multiplication and the convolution theorem):

im = np.mean(misc.imread('../images/mandrill.jpg'), axis=2) 
print(im.shape)
# (224, 225)
gauss_kernel = np.outer(signal.gaussian(11, 3), signal.gaussian(11, 3)) # 2D Gaussian kernel of size 11x11 with σ = 3
im_blurred = signal.fftconvolve(im, gauss_kernel, mode='same')
fig, (ax_original, ax_kernel, ax_blurred) = pylab.subplots(1, 3, figsize=(20,8))
ax_original.imshow(im, cmap='gray')
ax_original.set_title('Original', size=20)
ax_original.set_axis_off()
ax_kernel.imshow(gauss_kernel)
ax_kernel.set_title('Gaussian kernel', size=15)
ax_kernel.set_axis_off()
ax_blurred.imshow(im_blurred, cmap='gray')
ax_blurred.set_title('Blurred', size=20)
ax_blurred.set_axis_off()
fig.show()

The following screenshot shows the output of the preceding code block in the spatial domain:

The following code block shows how to plot the original and the blurred image spectrum after convolution:

import scipy.fftpack as fftpack
F1 = fftpack.fft2((im).astype(float))
F2 = fftpack.fftshift( F1 )
pylab.figure(figsize=(15,8))
pylab.subplot(1,2,1), pylab.imshow( (20*np.log10( 0.1 + F2)).astype(int), cmap=pylab.cm.gray)
pylab.title('Original Image Spectrum', size=20)
F1 = fftpack.fft2((im_blurred).astype(float))
F2 = fftpack.fftshift( F1 )
pylab.subplot(1,2,2), pylab.imshow( (20*np.log10( 0.1 + F2)).astype(int), cmap=pylab.cm.gray)
pylab.title('Blurred Image Spectrum', size=20)
pylab.show()

The following screenshot shows the output of the preceding code block in the frequency domain:

The same code block as the previous one is used to compute a blurred version of the Victoria Memorial Hall image, but this time the fftconvolve() function is invoked with a larger (50 x 50) Gaussian kernel with σ=10. The following screenshot shows the output of the same code with the new image and the modified kernel: