# Sigma computation in getGaussianKernel

**URL:** <https://forum.opencv.org/t/sigma-computation-in-getgaussiankernel/9876>\
**Category:** Uncategorized\
**Tags:** imgproc\
**Created:** [August 17, 2022, 9:57am UTC](https://forum.opencv.org/t/sigma-computation-in-getgaussiankernel/9876 "2022-08-17T09:57:15Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![pabloduque0](https://avatars.discourse-cdn.com/v4/letter/p/54ee81/32.png) [@pabloduque0](https://forum.opencv.org/u/pabloduque0)\
**Post date:** [August 17, 2022, 9:57am UTC](https://forum.opencv.org/t/sigma-computation-in-getgaussiankernel/9876/1 "2022-08-17T09:57:15Z")

</div>

Hello there!

I was reading the `getGaussianKernel` [documentation](https://docs.opencv.org/3.1.0/d4/d86/group __imgproc__ filter.html#gac05a120c1ae92a6060dd0db190a61afa) it mentiones the following about how sigma is computed when given a non positive value:

> sigma: Gaussian standard deviation. If it is non-positive, it is computed from ksize as `sigma = 0.3*((ksize-1)*0.5 - 1) + 0.8`

I was wondering where such computation is derived from.

Thanks in advance!

---

<div class="post-metadata">

**Author:** ![crackwitz](https://sea2.discourse-cdn.com/flex020/user_avatar/forum.opencv.org/crackwitz/32/14_2.png) [@crackwitz](https://forum.opencv.org/u/crackwitz)\
**Post date:** [August 17, 2022, 10:23am UTC](https://forum.opencv.org/t/sigma-computation-in-getgaussiankernel/9876/2 "2022-08-17T10:23:56Z")

</div>

IIRC that’s figuring _about_ 3 sigma in the kernel size, with some fiddling for rounding reasons.

a gaussian distribution has 99.7% of its probability mass within 3 sigma. that’s good enough for 8 bits of precision.

plug some values in, consider the result.

---

<div class="post-metadata">

**Author:** ![pabloduque0](https://avatars.discourse-cdn.com/v4/letter/p/54ee81/32.png) [@pabloduque0](https://forum.opencv.org/u/pabloduque0)\
**Post date:** [August 18, 2022, 11:36am UTC](https://forum.opencv.org/t/sigma-computation-in-getgaussiankernel/9876/3 "2022-08-18T11:36:23Z")

</div>

Yes, thanks, that makes sense. I was not sure if the fiddling part had a different reason behind it or not.

Thanks a lot!

---

<div class="post-metadata">

**Author:** ![pabloduque0](https://avatars.discourse-cdn.com/v4/letter/p/54ee81/32.png) [@pabloduque0](https://forum.opencv.org/u/pabloduque0)\
**Post date:** [August 24, 2022, 3:03pm UTC](https://forum.opencv.org/t/sigma-computation-in-getgaussiankernel/9876/4 "2022-08-24T15:03:45Z")

</div>

Posted something else but realised the calculation was for sigma, not for ksize. Thanks again!