So from the graphical introduction to convolution, we find that we have shifted the system response function, we have scaled it according to the x-ray intensity, and we have summed all these scaled responses together - to repeat -shift, scale, and add.
If the input function is f(x,y), a 2-D function, and the system response is h(x,y), then the mathematical expression for shifting to point (x',y') is
and the scaling by the input function intensity is
To get the output image g(x,y), we sum up over all points (x',y'):
This is the mathematical definition of the process called "convolution." Note one more time that both graphically and mathematically we have shifted the system response function h(x,y), scaled it by the x-ray intensity function f(x,y), and added (integrated) to produce the output response function g(x,y). Note also that the system response function h(x,y) is the system response to a delta function (or very narrow) input.
At the risk of boring you to tears, you'll note that the system response function h(t) has been shifted to time t', scaled by the fluctuation of the input f(t) at that time, and added to produce the output time activity response g(t).
Next (Fourier Transform, part 1)
Previous (convolution introduction)
Douglas J. Wagenaar, Ph.D., wagenaar@nucmed.bih.harvard.edu