1
$\begingroup$

So I've been really confused about the act of flipping a kernel in image processing. Is there any intuitive way to think about it or does it only help computing efficiently and maintain the mathematical symmetry?
Do we necessarily flip a kernel in image processing? (say the kernel is not symmetrical). Do we only flip it because we'll put it in a convolution later? Thanks in advance!

$\endgroup$
1
  • $\begingroup$ Virtually all kernels used in image processing are symmetric or antisymmetric, and we don't really care about the flipping. $\endgroup$
    – user16034
    Commented Dec 12, 2022 at 21:10

1 Answer 1

1
$\begingroup$

Flipping the kernel comes from the pure-math context of a convolution, where it is a natural thing to do. See this part of 3b1b's video on convolutions where he gives a great intuitive answer to your question.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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