Thread Subject: Image convolution, Fourier

Hi all,

I?m trying to get into the BLindDeconvolution of the
ImageProcessingToolbox and to do so I convolve a typical
image with a known gaussian PSF to see then whether the
algorithm returns a correct reconstructed PSF and image.
The problem is, that when I convolve the image using:

h = fspecial('gaussian',7,5);
Blurred=conv2(image,h);

I always the this weird mesh grid in the Fouriertransform of
the image as shown in this image with a PSF of 13 pixels:
http://img153.imageshack.us/my.php?image=dcvpsf13wi4.jpg
Another example with a 5 pixel PSF:
http://img501.imageshack.us/my.php?image=dcvpsf5vl6.jpg

The same occurs if I convolve the image using the imfilter
function. Is this periodic grid in the Blurred Fourier
(original image convolved with true PSF and it?s frequency)
an artefact or is there anything else I?m not aware of? I
could be wrong on this, but I?m pretty sure the grid
shouldn?t be there. Unfortunately it?s independent of the
image and just changes it?s appearance slightly with the PSF
applied. Any idea, advices, hints what went wrong??

Many thanks in advance and best regards,

Steffen

No one with an idea?
Any hint is much appreciated!!

Thanks!
Steffen

My last response on that post...
I take the silence that no one has ever encountered this
particular problem. Could eventually one of the matlab
officials comment the grid effect due to the convolution?

"Steffen " <rileksn@gmail.com> wrote in message
<g0cd03$kmk$1@fred.mathworks.com>...
> My last response on that post...
> I take the silence that no one has ever encountered this
> particular problem. Could eventually one of the matlab
> officials comment the grid effect due to the convolution?
>
>

The grid you see is related to the fact that your PSF is
windowed. In other words, your PSF is a Gaussian function
multiplied by a rectangular window. In the frequency
domain, that corresponds to the convolution of the Fourier
transform of the Gaussian with the Fourier transform of the
rectangular window. Your smaller PSF had a wider "grid" in
the frequency domain because of the relationship between the
Fourier transform of a rectangular window and that window's
width.

Also, for future reference, if you really want a response
from "matlab officials," you should contact The MathWorks
directly, possibly via www.mathworks.com/support.

Steve