Take the 2-minute tour ×
Signal Processing Stack Exchange is a question and answer site for practitioners of the art and science of signal, image and video processing. It's 100% free, no registration required.

I'm reading Computer Vision: Algorithms and Applications wich is available online as a PDF and on Chapter 3.2 page 111 it introduces convolution operators for images giving the formulation:

$g(i, j) = \Sigma_{k,l} f(i − k, j − l)h(k, l) = \Sigma_{k,l}f(k, l)h(i − k, j − l)$

since $(i,j)$ are pixel coordinates $f$ is the pixel intensity function for the image we're convolving and $h$ is the function for the convolution matrix it makes sense to me that $g(i,j)$ is equal to the first sum expression but I can't wrap my head around how this is equivalent to the second sum expression. If someone could explain this I would really appreciate it.

share|improve this question
3  
It's just a change of variables. You could replace k by e.g. k+5 or in this case i-k and the sum would be the same. You're just summing in a different order. –  nikie Jan 9 '13 at 7:34
add comment

1 Answer

up vote 6 down vote accepted

Te elaborate on nikie's comment that it is just a change of variables, consider the following.

Suppose that the original image is a single pixel of intensity $1$ at point $(0,0)$ and all other pixels have intensity $0$. Then, the result of the convolution of this image is $h(i,j)$, that is, this single pixel in the original image produces pixel intensity $h(i,j)$ at the point $(i,j)$. The actual image is a whole bunch of pixels, though, at different points, and the superposition principle says that the resulting pixel intensity $g(i,j)$ at point $(i,j)$ is the sum of of the pixel intensities produced by the various pixels in the original image. There are two ways of calculating $g(i,j)$. Let's go through the pixels in the original image in order of their distance from the point $(i,j)$. Well, the point $(i-k,j-l)$ is at distance $(k,l)$ from $(i,j)$ and the pixel intensity $f(i-k,j-l)$ there produces intensity $f(i-k,j-l)h(k,l)$ at $(i,j)$. This gives the first formula $$g(i,j) = \sum_k\sum_l f(i-k,j-l)h(k,l).$$ Alternatively, we can go through the pixels in natural order and ask what intensity the pixel $f(k,l)$ produces at the point $(i,j)$. Well, the distance is $(i-k,j-l)$ and so the intensity is $f(k,l)h(i-k,j-l)$ giving $$g(i,j) = \sum_k\sum_l f(k,l)h(i-k,j-l).$$

share|improve this answer
    
A great intuitive answer, thanks. –  Keith Jan 9 '13 at 23:22
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.