Afer some experiments, it seems to me that:

the integral (in the entire domain) of a convolution $(f*s)(x)$, of a filter function $f(x)$ with a a signal function $s(x)$, multiplied by a weight function $w(x)$,

$$ \int (f*s)(x) \cdot w(x) \, dx $$

, when all functions are integrable,

is equal to the integral of the signal function $s(x)$ weighted by the cross-correlation $(f\star w)(x)$ of the filter function $f(x)$ with the weight function $w(x)$,

$$ \int (f\star w)(x) \cdot s(x) \, dx $$

Is that right?

If it is, what is the proof?

If it is not, is there something similar that let one merge the filter and the weight functions to create a new weight function?


If I am not mistaken, this is just changing the order of integration (which has to be justified by integrability conditions): \begin{align} \int (f*s)(x)w(x) dx & = \int\int f(y-x)s(y)dy\, w(x)dx\\ & = \int\int f(y-x)w(x)dx\, s(y)dy\\ & = \int (f\star w)(y)s(y)dy. \end{align}

| cite | improve this answer | |
  • $\begingroup$ Thanks! I was almost reaching that proof, but your answer has a much cleaner formulation. $\endgroup$ – Cristóvão D. Sousa Oct 29 at 15:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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